DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. PCT/CN2020/082103, filed on 03/30/2020.
Response to Amendment
The Examiner thanks the applicant for the remarks, edits and arguments.
Regarding Tittle Objection
Applicant Remarks:
The applicant has amended the title and believes it is more descriptive which complies with issues addressed by the examiner.
Examiner Response:
The examiner has considered the title and believes it is more descriptive towards the core concept of the invention. The title objection has been withdrawn by the examiner.
Regarding Claim Rejections – 35 U.S.C. 112(b)
Applicant Remarks:
The applicant has made amendments to the claims to provide more background and meaning to the words and modules recited in the amended claims. The applicant believes that the current amended claims do not recite indefinite subject matter and disagrees with the examiner. The applicant states that the limitation, “the at least one digital device is configured to extract blocks of partial data” as an example. The applicant believes that the specification provides sufficient support of how the blocks of partial data can be extracted from a single item. Further the applicant states that the drawings also provide support for the claimed subject matter.
Next, the applicant argues that the claims are not designed to be interpreted to invoke 112(f), and that, arguendo, if the claims invoke 112(f) interpretation, the specification provides sufficient support for the given amended claim limitations. Further the applicant states that, in the 112(f) interpretation section of the office action, the examiner has segmented and partially quoted sections of limitations and has failed to consider limitations as a whole.
Next, the applicant argues that the limitations, “each of said respective different processing modules is configured to apply a convolution operation to its received block partial data from said fist item of content” and “a plurality of processing modules configured to receive respective blocks of partial data from said at least one data-source device.” Also do not recite indefinite subject matter. The applicant believes the processing module is sufficiently disclosed in the specification and that the convolution operation on the received data is also sufficiently disclosed in the specification and drawings.
Next, the applicant argues that the features in “the merging module is configured to construct a global model of the artificial intelligence module” is also another partial quote from the claims and the section the examiner has noted fails to recite the specific descriptive elements of this limitation. The applicant states that the process disclosed in the complete limitation is sufficiently disclosed in the specification and the supported by the drawings as well.
For the reasons stated above and, reasons stated in the remarks, the applicant believes that the amended claims no longer recite indefinite concepts, even if the claims are interpreted under 112(f), there is sufficient evidence to support the claimed subject matter. Therefore, the applicant bevies the rejection under 112(b) should be withdrawn.
Examiner Response:
The examiner has considered the amendments and the remarks from the applicant. It is noted that claim 8 was amended to recite a merging “module” instead of “unit”. Because of this amendment, this limitation of claim 8 is no longer considered indefinite under 112(b), However, because 112(f) is invoked by the language used in the claims, claim 8 still recites indefinite subject matter because the claims and specification fail to provide adequate structure to the claimed “merging module”. After reading the submitted remarks, the examiner would like to explain why the claims are interpreted using 35 U.S.C. 112(f) and how that relates to 35 U.S.C. 112(b) and why the claims limitations stated by the examier are determined to be indefinite under 35 U.S.C. 112(b).
Per the MPEP 2181(I), examiners are required evaluate claims and give the broadest reasonable interpretation of the claims and, “In determining the BRI, examiners should establish the meaning of each claim term consistent with the specification as it would be interpreted by one of ordinary skill in the art, including identifying and construing functional claim limitations. If a claim limitation recites a term and associated functional language, the examiner should determine whether the claim limitation invokes 35 U.S.C. 112(f). Application of 35 U.S.C. 112(f) is driven by the claim language, not by applicant’s intent or mere statements to the contrary included in the specification or made during prosecution.” (Emphasis added). Per this section of the MPEP, the examiner must evaluate claims and see if the invoke 112(f) and use “means-plus-function language” regardless if that was the applicant’s intent or not. Further, the MPEP provides a 3-Prong analysis for examiners to apply to claims where “means-plus-function language” might be used to see if a claim invokes 35 U.S.C. 112(f). This test is found at MPEP 2181(I) and states:
“(A) the claim limitation uses the term "means" or "step" or a term used as a substitute for "means" that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term "means" or "step" or the generic placeholder is modified by functional language, typically, but not always linked by the transition word "for" (e.g., "means for") or another linking word or phrase, such as "configured to" or "so that"; and
(C) the term "means" or "step" or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.”
Looking at the amended claims, the examiner has noted claims that require this analysis. For example, Claim 1 recites, “the at least one digital device is configured to extract blocks of partial data from a first item of content …”(Emphasis added). The remaining components of this claim limitation are not relevant for this test because the claim fails to recite any further structure of the “digital device” and is merely reciting a function of/for the “digital device”. Using the analysis in the MPEP, the examiner believes this limitation invokes 112(f) interpretation based on the 3-Prong analysis. Step (A) is met because the claim recites a “means” that is a generic placeholder, which is “digital device”. Step (B) is met because the generic place holder term is modified by using functional language, which is “configured to”. Finally, Step(C) is met because the claim fails to recite sufficient structure of the generic place holder, which is the “digital device”. Therefore, because of this example, and the examples listed in the previous office action, the examiner has determined that the claims do recite “means-plus-function language” and the claims, as they are currently written, invoke a 35 U.S.C. 112(f). The examiner would like to note that invoking 35 U.S.C. 112(f) does not constitute as a rejection under 35 U.S.C. 112(f), a rejection cannot be made under 35 U.S.C. 112(f). Invoking this means that the examiner must interpret the claims by using the specification to assist in directly disclosing the structure of the claimed objects in the claims such as, “digital device”, “merging module” and “processing modules”. To do this the specification is considered to see if it provides sufficient structure of the claimed subject matter in question. If the specification cannot adequately disclose the structure of the claimed subject matter in question, then the claimed limitation would be determined to recite indefinite subject matter as defined in 112(b). This is explained in the MPEP 2181(II), “35 U.S.C. 112(f) states that a claim limitation expressed in means- (or step-) plus-function language "shall be construed to cover the corresponding structure…described in the specification and equivalents thereof." "If one employs means plus function language in a claim, one must set forth in the specification an adequate disclosure showing what is meant by that language. If an applicant fails to set forth an adequate disclosure, the applicant has in effect failed to particularly point out and distinctly claim the invention as required by the 35 U.S.C. 112(b) [or the second paragraph of pre-AIA section 112]." In re Donaldson Co., 16 F.3d 1189, 1195, 29 USPQ2d 1845, 1850 (Fed. Cir. 1994) (en banc).” Taking all of this into consideration the examiner has evaluated the claims and the specification and was unable to find adequate structure for the claimed objects “digital device”, “merging module” and “Processing modules”. The examiner would also like to note that the functions of these elements is disclosed in the specification, however the structure of these elements is missing or it is unclear in the specification where the structure of these elements is disclosed. Further, the remarks also fail to cite a location in the specification where the adequate description of the structure of the “digital device”, “merging module” and/or “processing modules”. Therefore, the examiner believes the amended claims still recite indefinite language and fails to particularly point and distinctly claim the invention as required by the 35 U.S.C. 112(b).
Next, the applicant argues that the “the at least one digital devices is configured to extract blocks of partial data” is not indefinite and the “Applicant notes that the specification provides extensive discussion of various ways in which blocks of partial data can be extracted from a single item of content. In particular, exemplary approaches for dividing an item of content into portions for distribution are depicted and described with respect to Figs. 4A and 4B.”. The examiner would like to point again to the above reasons for invoking 35 U.S.C. 112(f) interpretation. This claim limitation is regarded as indefinite not because the block of partial data or its functions, but this claim is regarded as indefinite because the structure of the “digital device” is not sufficiently disclosed in the specification using the 112(f) interpretation. The applicant has failed to provided a citation of the specification which adequately discloses the structure of the “digital device”. Further the applicant has not amended this claim to not recite “means-plus-function language”, therefore this claim limitation is still rejected under 35 USC 112(b).
Next, the applicant argues that the examiner has used partial quotes from the claims when making rejection of claims. The applicant argues that this omits key elements of the claim limitation in question. As stated above the claims have been interpreted to invoke 35 U.S.C. 112(f). The examiner has evaluated claims that in particular recite the “mean-plus-function language”. Therefore, the key elements of these claim to evaluate are the “means” and the modify functional language. For example, claim 1 recites, “the at least one digital device is configured to extract blocks of partial data from a first item of content and distribute the blocks of partial data to respective different processing modules in the intermediate processing system;”. As stated above the generic placeholder for the “means” is cited to be the “digital device” and the modifying language is “configured to”. The next part of the claim limitation recites the function of the digital device which is, “extract blocks of partial data from a first item of content and distribute the blocks of partial data to respective different processing modules in the intermediate processing system;”. The last section of this claim limitation is merely reciting what the “digital device” does and fails to recite adequate structure of the “digital device”. Therefore, the indefinite subject matter is not the function the object performs, but the indefinite subject matter is the “digital device” itself. Finally, the remaining function of the claim limitation is still considered as a whole when evaluating claims, however the examiner has shortened the quoted section in the office action to highlight the key elements of the claim limitation being rejected.
Next, the applicant argues that the claim limitations, 1. "each of said respective different processing modules is configured to apply a convolution operation to its received block of partial data from said first item of content " and 2. "a plurality of processing modules configured to receive respective blocks of partial data from said at least one data-source device," is not indefinite subject matter. The examiner would like to point to the stated reasons above for invoking 112(f). The examiner has reviewed these claims and believes both limitations recite “means-plus-function” language. After applying the 3-prong analysis the examiner believes these limitations invoked 112(f). Limitation 1. recites "each of said respective different processing modules is configured to apply a convolution operation to its received block of partial data from said first item of content” (emphasis added). Step (A) is met because the claim recites a generic placeholder, which is “processing modules”. Step (B) is met because the claim recites modifying language, which is “configured to”. Finally, Step (C) is met because the claim fails to recite sufficient structure of the “processing modules”. This analysis is also applied to limitation 2. The generic place holder (Step(A)) in limitation 2. is the “processing modules” which is followed by modifying functional language, (Step(B)) which is “configured to”. Finally, using Step(C), the limitation fails to further define the structure of the “processing modules”. Because of this analysis of both limitations, the specification must be used to ensure adequate disclosure of the structure of the “processing modules”. After reviewing the specification, the examiner was unable to find a citation which adequately discloses the structure of the “processing modules”. Therefore, the examiner believes the amended claim limitation still recite indefinite language and fails to particularly point and distinctly claim the invention as required by the 35 U.S.C. 112(b). Further the examiner would also like to note that the claim limitations as a whole are always considered and the examiner does not only evaluate partial quotes or limitations. However, as stated above, the rejection under 35 U.S.C. 112(b) is not for the functions stated in the limitations, the rejection is for the object disclosed in the claims that is performing the functions listed, i.e. “processing modules” is the key issue, not the function it performs.
Next, the applicant argues that the examiner has used a partial quote of the limitation and notes that the remaining elements of the limitation should be regarded. As stated above, the limitation in question has been evaluated to invoke 35 USC 112(f). the claim recites "the merging module is configured to construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said respective different processing modules.". The examiner would like to point to the above reasons for invoking 112(f) on this claim and the 3-Prong analysis is applied to this limitation. Step (A) is met because the limitation discloses a generic placeholder, which is “Merging module”. Step (B) is met because the claim recites modifying functional language, which is “configured to”. Finally, Step (C) is met because the rest of the claim recites the function of the “merging module” and fails to recite the structure of the “merging module”. Since 35 USC 112(f) in invoked, the examiner is required to evaluate the specification to see if adequate disclosure of the “merging module” is present. After review of the specification, the examiner was unable to locate a citation which adequately disclosed the structure of the “merging module”. The applicant has failed to provide a citation in the specification which adequately discloses the structure of the “merging module”. Further the applicant has not amended this claim to not recite “means-plus-function language”, therefore this claim limitation is still rejected under 35 USC 112(b). As for using the partial quote, the examiner was highlighting the key elements of the claim which discloses the indefinite subject matter. The remaining elements of the claim recite a function that is performed by the “margining module” and fails to recite further structure of the “merging module”. Therefore, the key elements of this claim is the "the merging module is configured to construct a global model of the artificial intelligence module" The claim limitation as a whole is always considered when evaluating claims.
Finally, for the reasons stated above, and in the previous office action, the claims are interpreted to invoke 35 USC 112(f). This interpretation is used based on the language presented in the claims in accordance to MPEP 2181, regardless if the applicant intended to invoke 35 U.S.C. 112(f). Since this interpretation is being used for these claims, the specification must provide adequate information to the structure of the claimed objects listed in the claims. Failure to do so would result in a 35 U.S.C. 112(b) rejection as failing to particularly point and distinctly claim the invention as required by the 35 U.S.C. 112(b). Finally, as stated above, the claim limitations and specification have been evaluated to see if they comply with 35 USC 112(b). For the reasons stated above and in the prior office action, the examiner believes the rejection under 35 USC 112(b) remains and is upheld.
Regarding Claim Rejections – 35 U.S.C. 102
Applicant Remarks:
The applicant argues that Ye fails to disclose each and every element of the amended claims. The applicant states that Ye fails to disclose a system which takes in a single data item and partitions the data item and then sends the different partitions of the single data item to different processing modules for parallel processing. The applicant believes that Ye does not disclose the use of a single data items being partitioned as claimed and Ye does not send partitions of the single data items to different processing units for parallel processing. The applicant believes that Ye discloses a process of taking in data into a federated system and each device in the system contains its own data sets. The applicant believes that Ye is fundamentally different that the claimed subject matter. Because of the key differences the applicant believes that Ye fails to anticipate the amended claims and, because of claim dependency, Vepakomma, would also fail to teach the elements missing from Ye. Therefore, the applicant believes that the rejection under 35 U.S.C. 103 is also incorrect. For the reasons stated above and in the submitted remarks, the applicant believes that the rejection under 35 U.S.C. 102 should be withdrawn.
Examiner Response:
The applicant argues that Ye is fundamentally different than the claimed subject matter. The applicant also states that, “a single source device takes one item of content, and sends fragments of that one item to different processing modules.” And that Ye fails to disclose, “a digital device that extracts blocks of partial data from a single, common item of content, and distributes those blocks to different processing modules.”. The examiner would like to point to the current amended Claim 1 which states, explicitly, “the at least one digital device is configured to extract blocks of partial data from a first item of content and distribute the blocks of partial data to respective different processing modules in the intermediate processing system;” (emphasis added). Claim 1 explicitly states that the partial data being extracted is from a “first item” not “a singular common item of content.” Therefore, using the broadest reasonable interpretation and in light of the 112(f) interpretation, the examiner has interpreted the “first item” as data from an item that is first in a set of items. Ye teaches a federated system that uses sensor data from multiple sources including cameras. The system in Ye is able to take data from an item, i.e. a single vehicle in a set of vehicles, and export the data for further processing.
Finally, the examiner has evaluated the claims under Ye and 35 U.S.C. 102 and believes that Ye is able to disclose or teach the broadest reasonable interpretation of most but not each and every element of the amended claims. A complete search is completed after each and every amendment. After performing this search, the examiner was unable to locate a single prior art that is able to anticipate the claims as disclosed, this includes Ye. For this reason, the examiner believes the current amended claims comply with 35 U.S.C. 102 and the examiner has withdrawn this rejection.
Regarding Claim Rejections – 35 U.S.C. 103
Applicant Remarks:
The applicant argues that the art Ye fails to disclose the each and every element of the amended claims. Therefore, because of the claim dependency, the examiner would need to rely on Vepakomma teach the missing elements. Since Vepakomma cannot teach the missing elements of the independent claims, the combination of Ye and Vepakomma would fail to teach the dependent claims. Further, the applicant states that the architecture of machine learning models in Ye and Vepakomma are fundamentally different than the claimed invention.
Next, the applicant argues that a person of regular skill in the art would not reasonably combine or consider using Ye and Vepakomma together to disclose the current invention. Ye discloses a federated learning system which is able to take data and model outputs from different Vepakomma in a federated system. Vepakomma discloses a system that uses split learning and is able to take model outputs from different devices and processes the data using a global server. Both of these articles are fundamentally different and the applicant believes it would not be obvious to a person of ordinary skill in the art to combine these arts and/or combine them in a way to anticipate the claimed invention.
Next, applicant argues that Vepakomma also fails to disclose the use of single item of digital content which is partitioned and the partitions are sent to different processing modules for further processing. The applicant argues that the split architecture of Vepakomma does not split or partition a single data item to send to the server, instead the applicant believes that the devices in this system contain their own training data and send outputs, generated from their own separate data stores, of the models to the sever for further processing.
Finally, the applicant argues that the architectures of Ye and Vepakomma are fundamentally different that the claimed subject matter and that no combination of Ye and Vepakomma would be able to disclose or teach each and every element which is claimed. Therefore, the applicant believes the claims disclose a novel process that is different form the proposed art and the rejection under 35 U.S.C. 103 should be withdrawn.
Examiner Response:
The applicant argues it would not be obvious for an ordinary person of the art to combine the elements of Ye and Vepakomma. The examiner believes that it would have been obvious for an ordinary person in the art to combine Ye and Vepakomma to disclose the claimed invention. Ye does disclose a standard federated learning model where vehicles are connected to a central server. It is stated in the specification in the claims recite a form of federated learning. Next, Vepakomma discloses different forms of splitting neural networks and methods to preserve privacy of clients and users. Both of these methods use a form of distributed learning across multiple devices. A person of ordinary skill would have motivation to apply concepts of federated learning with concepts of split learning. Further, Vepakomma has submitted other NPL called “Detailed comparison of communication efficiency of split learning and federated learning.” Where the same author discloses a comparison of the two types of distributed learning. All of these articles were available to the public prior to the effective filing date of this application and a person interested in learning about distributed learning would have been able to locate and read these articles.
Next, the applicant argues that the vertical partitioning of data in Vepakomma does not disclose the use of partitioning a single item of digital content. The examiner would like to point to the claims which recite the extraction of data from a “first item”. As stated above, the “first item” is not explicitly defined in the claims or the specification. The claims fail to disclose the “partitioning of a single item of digital content” and instead disclose a extracting partial data from a “first item of content”. While evaluating the claims the broadest reasonable interpretation is used. The term “first item of content” has been interpreted as a first item in a set of items which contain content of some form. Vepakomma discloses a distributed system where the data, which is considered items of content, are initially evaluated by the client devices and kept separate from other clients. Further the applicant states that Vepakomma does not transmit “raw data” to an intermediate server for processing. The examiner would like to not that the claims and the specification fails to recite the phrase “raw data”. Therefore, the examiner did not attempt to search for this concept.
Finally, the applicant argues that Ye and Vepakomma fail to disclose the independent claims and as a result would fail to teach each and every element of the claims. After each amendment the examiner must perform a complete and through search and reconsider the claims and changed meanings. As stated above, the examiner has found the applicants arguments against Ye and the rejection under 35 U.S.C. 102 to be persuasive, however, the examiner believes that Ye does still teach some elements of the claims. After a search was completed, the examiner did find new art that, in combination of the previously presented art, is able to disclose or teach each and every element of the claimed subject matter. Therefore, the examiner believes that the rejection under 35 U.S.C. 103 is to be upheld, see 103 rejection below.
Regarding Claim Rejections – 35 U.S.C. 101
Applicant Remarks:
The applicant has made amendments to the claims and believes they do not recite abstract ideas and the claims, as written, would be considered patent eligible under 35 U.S.C. 101. Further the applicant believes that the claim limitations the examiner has marked as abstract ideas are not abstract and cannot be performed in a human mind. Further, the applicant remarks that the examiner has failed to provide sufficient reasons or evidence as to why the stated limitations recite an abstract idea. The applicant believes the examiner has failed to state explicit reasoning on why limitations are considered abstract and the believes that the examiner has broadly interpreted the claims to recite a process of evaluation and observation of data.
Next, the applicant argues that the system disclosed in the claims requires the use of complex systems in order to process and execute the claimed method. The proposed system is connected to multiple devices in a federated network which would be impossible to create or imagine solely in a human mind with the assistance of pen and paper. The applicant states that the claimed subject matter discloses novel architecture which allows of different devices of lower computational complexity to evaluate and process data in parallel. The applicant states that this process is consistent with recent PTAB decisions and that the subject matter recites sufficient evidence of an improvement to machine learning technology. The applicant believes that the claimed subject matter in light of the specification, recites an improvement to a technical field and therefore the claims as a whole would be integrate an exception into a practical applicant. Therefore, the claimed process would be considered patent eligible under Step 2A Prong 2.
Finally, the applicant states that claim 1 has been amended to further clarify the claim and the processes stated in the claim cannot be performed by human mind with the assistance of pen and paper. For the reasons stated above and in the submitted remarks, the applicant believes the current amended claims comply with 35 U.S.C. 101 and the rejection under this section should be withdrawn.
Examiner Response:
The examiner has evaluated and considered the current amended claims and the remarks from the applicant. While reviewing the claims, the examiner noted the amendments made which further defines the intermediate processing system and the merging module. These amendments allow the broadest reasonable interpretation to require the use of a computer or some form of processing system outside of a human mind. Therefore, the examiner has found the applicants argument is persuasive and these limitations would not be considered abstract ideas. However, the claimed structure of the processing device and merging module while using the broadest reasonable interpretation, would lead a person of ordinary skill in the art to recognize a “merging module” and a “intermediate processing device” as generic computing devices such as personal computers, servers or handheld computing devices. This is important when looking at the limitations, 1. “each of said respective different processing modules receiving a block of partial data from said first item of content is configured to: apply processing to the block of partial data to generate a partial model for an artificial intelligence module, and” and 2. “the merging module is configured to construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said respective different processing modules.”. The broadest reasonable interpretation of the claims as a whole is used when evaluating if a claim recites a mental process. Limitation 1. discloses a processing module that will receive data from another source and process the data to generate a partial model. The examiner believes that this limitation still recites an abstract idea of evaluating and judging information. Per the MPEP 2106.04(a)(2)(III)(C) states, “In evaluating whether a claim that requires a computer recites a mental process, examiners should carefully consider the broadest reasonable interpretation of the claim in light of the specification. For instance, examiners should review the specification to determine if the claimed invention is described as a concept that is performed in the human mind and applicant is merely claiming that concept performed 1) on a generic computer, or 2) in a computer environment, or 3) is merely using a computer as a tool to perform the concept. In these situations, the claim is considered to recite a mental process.”. The examiner believes the first limitation discloses an abstract idea, “apply processing to the block of partial data to generate a partial model for an artificial intelligence module”. This is an abstract idea because a human is able to process data and generate a judgment or output from the evaluated data. For example, a human is able to read a computer screen, extract information from that screen and generate a partial or complete model based on the information evaluated. Further this process is using a generic computing device, or “each of said respective different processing modules”, as a tool to complete the abstract idea of processing data. Next regarding the second limitation, recites a process of using a merging module to process the partial models. The examiner believes that this core concept is a process of evaluating and judging data. A human is able to evaluate data and from that evaluation a human can provide feedback or judgement to produce a complete model or outcome. As stated above the “merging module” is interpreted to be a generic computing device and therefore is merely a tool used to complete the abstract idea. Therefore, further processing of the claims is required per the Alice/Mayo test and the remaining limitations are evaluated under Step 2A prong 2.
Next, the examiner recognizes that the explanations given in the previous office action appear dry and without supporting evidence. The examiner did not intent to cause confusion or ambiguity while explaining reasons for rejections and believed the explanations given were sufficient. Moving forward, for better compact prosecuting, the examiner will explain interpretations, explanations of rejections, and comments in more detail. The examiner advises the applicant to contact either the examiner or the examiners SPE directly via phone call, phone numbers are listed in the conclusion, if the applicant believes the examiner is acting unfairly or punitively. Further the examiner would like to note that the applicant can schedule an interview with the examiner anytime to discuss the application and/or rejections made.
Next, The applicant argues that, even if the claims recite abstract ideas, the claims still recite improvements to machine learning models and a technical field. The examiner has noted that the specification recites possible improvements such as distributed parallel processing of data items and privacy preserving techniques. However, the examiner would like to point to MPEP 2106.05(a) which states, “If it is asserted that the invention improves upon conventional functioning of a computer, or upon conventional technology or technological processes, a technical explanation as to how to implement the invention should be present in the specification. That is, the disclosure must provide sufficient details such that one of ordinary skill in the art would recognize the claimed invention as providing an improvement. The specification need not explicitly set forth the improvement, but it must describe the invention such that the improvement would be apparent to one of ordinary skill in the art.” (emphasis added). The examiner has considered the specification and believes that a person of ordinary skill in the art would not be able to recognize the claimed invention as an improvement to parallel processing, privacy preserving techniques or machine learning. The specification discloses the invention and how it is implemented, however it is unclear where it states how this invention improves a technical field or technology. Pages 3 and 4 of the specification do disclose possible improvement of partitioning data, parallel processing and privacy preserving techniques, however these appear to be conventional and well understood concepts data partitioning, parallel processing and federated learning. Further the applicant has not provided any citation from the specification or the claims disclosing the technical improvement to any technical field or technology. The applicant has disclosed in the remarks that the invention is able to offload computational workloads to other devices in a network as an improvement. However, it appears to recite a novel concept of distributed computing networks and a person of ordinary skill in the art would be able to recognize this.
Next, applicant argues that, per recent updates to the MPEP, the specification clearly identifies an improvement to machine learning technology “by explaining how the model is updated while preserving the privacy of the source data in a novel manner.”. The examiner would like to note the above paragraph in regarding the specification failing to provide an improvement recognizable to a person of ordinary skill in the art. Next, the examiner would like note the specification fails to disclose a privacy preserving technique while updating a model. The specification does not contain the work “update” or “updating” and fails to disclose an model updating process using privacy preserving techniques. Therefore, the examiner disagrees with the applicant that the specification clearly identifies an improvement.
Finally, The examiner has reviewed the amended claims and has noted that the claims recite abstract concepts, and the remaining limitations fail to integrate judicial exceptions into practical application, and the specification and claims fail to recite a technical improvement to technology or a technical field. For these reasons and the reasons stated above, the examiner believes the current claims fail to comply with 35 U.S.C. 101 and the rejection under 35 USC 101 is upheld.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier.
Such claim limitations In Claim 1 are:
“the at least one digital device is configured to extract blocks of partial data from a first item of content and distribute the blocks of partial data to respective different processing modules in the intermediate processing system;”
“each of said respective different processing modules receiving a block of partial data from said first item of content is configured to:”
“the merging module is configured to construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said respective different processing modules.”
In claim 2:
“wherein said at least one digital device is configured to extract blocks of partial data from the first item of content by extraction of specific target features in the item of content.”
In claim 3:
“wherein said at least one digital device is configured to extract blocks of partial data from the first item of content by dividing the first item of content into blocks of partial data, the blocks having a specified size.”
In claim 4:
“wherein said at least one digital device is configured to transmit ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said respective different processing modules are configured to transmit to the merging module their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.”
In claim 5:
“wherein each of said respective different processing modules is configured to apply a convolution operation to its received block of partial data from said first item of content.”
In claim 6:
“wherein the merging module is configured to construct a global model of the artificial intelligence module by performing a series of pooling and convolution operations on the partial models received from said respective different processing modules, and by training a fully-connected neural network using the results of said pooling and convolution operations.”
In claim 8:
“a plurality of processing modules implemented on at least one server distinct from the server implementing the merging module, each of the respective plurality of processing modules configured to receive respective blocks of partial data from said at least one data-source device, said blocks of partial data being extracted from a common item of digital content;”
“wherein each of the respective different processing modules is configured to: apply processing to its respective received block of partial data to generate a partial model for an artificial intelligence module, and”
In claim 9:
“wherein said processing modules are configured to receive ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said processing modules are configured to transmit to the merging module of the AI-module training system their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.”
In claim 10:
“wherein each of said processing modules is configured to apply a convolution operation to its respective received block of partial data from said first item of content.”
In claim 11:
“wherein said digital device is configured to: extract blocks of partial data from a first item of content; and”
In claim 12:
“wherein said merging module is configured to: receive partial models from different processing modules m the intermediate processing system; and”
In claim 17:
“wherein each of the plurality of processing modules are configured to perform at least a first convolution operation of a convolution neural network algorithm to generate respective partial models for the artificial intelligence module.”
In claim 18:
“wherein the merging module is configured to perform subsequent steps of the convolution neural network algorithm not performed by the plurality of processing modules to construct the global model of the artificial intelligence module.”
Emphasis is added to the “means-plus-function” sections and the whole claim limitations are stated. Because these claim limitations are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, they are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have these limitations interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitations to avoid them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitations recites sufficient structure to perform the claimed function so as to avoid them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-6, 8-12, 17 and 18 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as failing to set forth the subject matter which the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the applicant regards as the invention.
Claim Limitations:
In claim 1:
“the at least one digital device is configured to extract blocks of partial data from a first item of content and distribute the blocks of partial data to respective different processing modules in the intermediate processing system;”
“each of said respective different processing modules receiving a block of partial data from said first item of content is configured to:”
“the merging module is configured to construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said respective different processing modules.”
In claim 2:
“wherein said at least one digital device is configured to extract blocks of partial data from the first item of content by extraction of specific target features in the item of content.”
In claim 3:
“wherein said at least one digital device is configured to extract blocks of partial data from the first item of content by dividing the first item of content into blocks of partial data, the blocks having a specified size.”
In claim 4:
“wherein said at least one digital device is configured to transmit ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said respective different processing modules are configured to transmit to the merging module their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.”
In claim 5:
“wherein each of said respective different processing modules is configured to apply a convolution operation to its received block of partial data from said first item of content.”
In claim 6:
“wherein the merging module is configured to construct a global model of the artificial intelligence module by performing a series of pooling and convolution operations on the partial models received from said respective different processing modules, and by training a fully-connected neural network using the results of said pooling and convolution operations.”
In claim 8:
“a plurality of processing modules implemented on at least one server distinct from the server implementing the merging module, each of the respective plurality of processing modules configured to receive respective blocks of partial data from said at least one data-source device, said blocks of partial data being extracted from a common item of digital content;”
“wherein each of the respective different processing modules is configured to: apply processing to its respective received block of partial data to generate a partial model for an artificial intelligence module, and”
In claim 9:
“wherein said processing modules are configured to receive ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said processing modules are configured to transmit to the merging module of the AI-module training system their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.”
In claim 10:
“wherein each of said processing modules is configured to apply a convolution operation to its respective received block of partial data from said first item of content.”
In claim 11:
“wherein said digital device is configured to: extract blocks of partial data from a first item of content; and”
In claim 12:
“wherein said merging module is configured to: receive partial models from different processing modules in the intermediate processing system; and”
In claim 17:
“wherein each of the plurality of processing modules are configured to perform at least a first convolution operation of a convolution neural network algorithm to generate respective partial models for the artificial intelligence module.”
In claim 18:
“wherein the merging module is configured to perform subsequent steps of the convolution neural network algorithm not performed by the plurality of processing modules to construct the global model of the artificial intelligence module.”
invokes 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. However, the written description fails to disclose the corresponding structure, material, or acts for performing the entire claimed function and to clearly link the structure, material, or acts to the function. The items that are indefinite are emphasized above. The disclosure of this claim is devoid of any structure that performs the function in the claim. This claim discloses an apparatus which does not further teach the structure which the functions are performed on. Therefore, the claim is indefinite and is rejected under 35 U.S.C. 112(b) or pre-AIA 35 U.S.C. 112, second paragraph.
Applicant may:
(a) Amend the claim so that the claim limitation will no longer be interpreted as a limitation under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph;
(b) Amend the written description of the specification such that it expressly recites what structure, material, or acts perform the entire claimed function, without introducing any new matter (35 U.S.C. 132(a)); or
(c) Amend the written description of the specification such that it clearly links the structure, material, or acts disclosed therein to the function recited in the claim, without introducing any new matter (35 U.S.C. 132(a)).
If applicant is of the opinion that the written description of the specification already implicitly or inherently discloses the corresponding structure, material, or acts and clearly links them to the function so that one of ordinary skill in the art would recognize what structure, material, or acts perform the claimed function, applicant should clarify the record by either:
(a) Amending the written description of the specification such that it expressly recites the corresponding structure, material, or acts for performing the claimed function and clearly links or associates the structure, material, or acts to the claimed function, without introducing any new matter (35 U.S.C. 132(a)); or
(b) Stating on the record what the corresponding structure, material, or acts, which are implicitly or inherently set forth in the written description of the specification, perform the claimed function. For more information, see 37 CFR 1.75(d) and MPEP §§ 608.01(o) and 2181.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-13 and 15-18 are rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis of the claims will follow the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (“2019 PEG”).
Claim 1
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Claim 1 recites, “An artificial-intelligence-module training system to train an AI module, the training system comprising:” therefore it is directed to the statutory category of a machine.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“the at least one digital device is configured to extract blocks of partial data from a first item of content and distribute the blocks of partial data to respective different processing modules in the intermediate processing system;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to observe data, either physical or digital, and extract data from that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“apply processing to the block of partial data to generate a partial model for an artificial intelligence module, and” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to take an observation of data and evaluate that data and further provide and opinion or judgement based on the evaluation. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“the merging module is configured to construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said respective different processing modules.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to combine data using different functions and is able to take parts of a model and combine them in a way to get a whole model. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “a source of training data, said source of training data comprising at least one digital device;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“an intermediate processing system comprising at least one server, the intermediate processing system comprising a plurality of processing modules, each of the plurality of processing modules implemented on said at least one server of the intermediate processing system; and” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“a merging module implemented on a server distinct from the at least one server of the intermediate processing system;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“each of said respective different processing modules receiving a block of partial data from said first item of content is configured to:” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“transmit the partial model to the merging module; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “a source of training data, said source of training data comprising at least one digital device;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“an intermediate processing system comprising at least one server, the intermediate processing system comprising a plurality of processing modules, each of the plurality of processing modules implemented on said at least one server of the intermediate processing system; and” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“a merging module implemented on a server distinct from the at least one server of the intermediate processing system;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“each of said respective different processing modules receiving a block of partial data from said first item of content is configured to:” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“transmit the partial model to the merging module; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 2
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“wherein said at least one digital device is configured to extract blocks of partial data from the first item of content by extraction of specific target features in the item of content.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use a computer to extract and evaluate data from a data source and locate specific target features in the data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim 3
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“wherein said at least one digital device is configured to extract blocks of partial data from the first item of content by dividing the first item of content into blocks of partial data, the blocks having a specified size.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from a data source using a generic computer. Further a human is able to observe data and partition data into smaller sections using well understood methods. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim 4
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein said at least one digital device is configured to transmit ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said respective different processing modules are configured to transmit to the merging module their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein said at least one digital device is configured to transmit ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said respective different processing modules are configured to transmit to the merging module their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 5
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“wherein each of said respective different processing modules is configured to apply a convolution operation to its received block of partial data from said first item of content.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to observe and apply functions to observed and evaluated data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim 6
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein the merging module is configured to construct a global model of the artificial intelligence module by performing a series of pooling and convolution operations on the partial models received from said respective different processing modules, and by training a fully-connected neural network using the results of said pooling and convolution operations.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein the merging module is configured to construct a global model of the artificial intelligence module by performing a series of pooling and convolution operations on the partial models received from said respective different processing modules, and by training a fully-connected neural network using the results of said pooling and convolution operations.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 7
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein the intermediate processing system comprises a set of servers implementing said processing modules.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein the intermediate processing system comprises a set of servers implementing said processing modules.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 8
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Claim 8 recites, “A processing system configured for use in an AI-module training system, said AI-module training system comprising at least one data-source device and a merging module implemented on a server, the processing system comprising:” therefore it is directed to the statutory category of a machine.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“wherein each of the respective different processing modules is configured to: apply processing to its respective received block of partial data to generate a partial model for an artificial intelligence module, and” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use generic computing devices to evaluate and observe data and apply judgements to generate a partial or whole model. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “a plurality of processing modules implemented on at least one server distinct from the server implementing the merging module, each of the respective plurality of processing modules configured to receive respective blocks of partial data from said at least one data-source device, said blocks of partial data being extracted from a common item of digital content;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“transmit the partial model to the merging module of the AI-module training system.” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “a plurality of processing modules implemented on at least one server distinct from the server implementing the merging module, each of the respective plurality of processing modules configured to receive respective blocks of partial data from said at least one data-source device, said blocks of partial data being extracted from a common item of digital content;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“transmit the partial model to the merging module of the AI-module training system.” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 9
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein said processing modules are configured to receive ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said processing modules are configured to transmit to the merging module of the AI-module training system their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein said processing modules are configured to receive ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said processing modules are configured to transmit to the merging module of the AI-module training system their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 10
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“wherein each of said processing modules is configured to apply a convolution operation to its respective received block of partial data from said first item of content.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses mathematical concept of utilizing a mathematical formula to perform calculations. The core elements of a convolutional neural network, such as applying a convolutional operation, require mathematical functions. A human is able to perform these functions using known mathematical concepts. This claim discloses a math operation and therefore is ineligible.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim 11
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“wherein said digital device is configured to: extract blocks of partial data from a first item of content; and” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use a generic computer to evaluate and observe data and a human is able to extract data from another data source. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “distribute the blocks of partial data to respective different processing modules in the intermediate processing system, so that each of said different processing modules may apply processing to the block of partial data to generate a partial model for the artificial intelligence module and transmit said partial model to a merging module in the artificial-intelligence-module training system.” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “distribute the blocks of partial data to respective different processing modules in the intermediate processing system, so that each of said different processing modules may apply processing to the block of partial data to generate a partial model for the artificial intelligence module and transmit said partial model to a merging module in the artificial-intelligence-module training system.” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 12
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said different processing modules.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to take data from different sources and observe the data and apply judgements or opinions to that data to generate partial or complete model. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein said merging module is configured to: receive partial models from different processing modules m the intermediate processing system; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein said merging module is configured to: receive partial models from different processing modules m the intermediate processing system; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 13
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Claim 13 recites, A computer-implemented method to train an artificial intelligence module, the method comprising:” therefore it is directed to the statutory category of a process.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“a) extracting blocks of partial data from a first item of content stored or acquired at a digital device,” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. The examiner has interpreted this limitation to recite the process of evaluating data using a generic computer as a tool. A human is able to use a computer as a tool to extract information, partial or complete, from a given data item. A human is able to evaluate stored data from one’s own mind or by extracting data from a tangible object such as using a computer monitor. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“c) applying processing to the block of partial data, by the respective different processing systems, to generate respective partial models for the artificial intelligence module;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. The examiner has interpreted this limitation to recite the process of evaluating data and applying modifications such as opinions judgements to observed data. A human is able to evaluate blocks of data and is able to generate partial or whole models based on the observed data. A human is able to use a computer as a tool to perform this action as well. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“e) constructing a global model of the artificial intelligence module by the performance of processing operations by the merging module, said processing operations including processing the partial models received from the respective different processing modules.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. The examiner has interpreted this limitation to recite the process of evaluating data and applying modifications such as opinions judgements to observed data. A human is able to use observed data or models and process or combine them to generate a new model. A human is able to use a computer as a tool to perform this action as well. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “b) distributing the blocks of partial data from the digital device to respective different processing modules in an intermediate processing system;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible. The examiner has used the broadest reasonable interpretation of this limitation and believes it recites a process of sending data, blocks of partial data, from one device, the digital device, to another, different processing modules in an intermediate processing system. This limitation recites a well understood process of transmitting data over a network and is common in the art.
“d) transmitting the partial models from the intermediate processing system to a merging module; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible. The examiner has used the broadest reasonable interpretation of this limitation and believes it recites a process of sending data, blocks of partial models, from one device, intermediate processing system, to another, a merging module. This limitation recites a well understood process of transmitting data over a network and is common in the art.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “b) distributing the blocks of partial data from the digital device to respective different processing modules in an intermediate processing system;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
The claim recites the additional elements, “d) transmitting the partial models from the intermediate processing system to a merging module; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 14 (Cancelled)
Claim 15
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above. This claim is written in dependent form and therefore it would fall under the same statutory category as claim 13, which is a process.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“A non-transitory computer-readable medium having stored thereon instructions which, when executed by a processor, cause the processor to carry out steps c) and d) of the method according to claim 13.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. The examiner interprets this claim to recite a process of executing the steps of claim 13 using a generic computer. As stated above, claim 13 discloses limitations which recite abstract concepts of evaluating and processing data. This limitation is merely applying an abstract idea, claim 13, on generic computer system, the non-transitory computer readable medium”. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim 16
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein no processing module of the plurality of processing modules receives the entire first item of content.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein no processing module of the plurality of processing modules receives the entire first item of content.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 17
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“wherein each of the plurality of processing modules are configured to perform at least a first convolution operation of a convolution neural network algorithm to generate respective partial models for the artificial intelligence module.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses mathematical concept of utilizing a mathematical formula to perform calculations. The core elements of a convolutional neural network, such as applying a convolutional operation, require mathematical functions. A human is able to perform these functions using known mathematical concepts. This claim discloses a math operation and therefore is ineligible.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim 18
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A machine, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The claim recites, inter alia:
“wherein the merging module is configured to perform subsequent steps of the convolution neural network algorithm not performed by the plurality of processing modules to construct the global model of the artificial intelligence module.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses mathematical concept of utilizing a mathematical formula to perform calculations. The core elements of a convolutional neural network, such as applying a convolutional operation, require mathematical functions. A human is able to perform these functions using known mathematical concepts. This claim discloses a math operation and therefore is ineligible.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-4, 8, 9, 11-13, 15, 16, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Lyu et al, (Lyu et al, “Fog-Embedded Deep Learning for the Internet of Things”, 2019, hereinafter “Lyu”) in view of Ye et al, (Ye et al, “Federated Learning in Vehicular Edge Computing: A Selective Model Aggregation Approach”, 2020, hereinafter “Ye”).
Regarding Claim 1, Lyu discloses, “An artificial-intelligence-module training system to train an AI module, the training system comprising:” (Algorithm 1: FPPDL Training Procedure, pp. 4209; This algorithm discloses a training method for training local and global federated learning models in a fog computing system. This algorithm discloses the process for the client, fog nods and the cloud server.)
“an intermediate processing system comprising at least one server, the intermediate processing system comprising a plurality of processing modules, each of the plurality of processing modules implemented on said at least one server of the intermediate processing system; and” (FPPDL Framework, pp. 4209; “We display a three-level Fog-embedded framework for deep learning in Fig. 1(d), which can support numerous end nodes at the bottom level, Fog nodes in the middle Fog level, and the Cloud at the top level. The Fog nodes have two functions: (i) as Fog servers with more powerful storage and processing capability than end nodes; (ii) as Fog edge nodes that can interact with heterogenous devices with different protocols [17].” This article discloses a system that uses a three-level framework to computer data. The system has a edge user able to generate or input data, an intermediate fog node able to perform processing of data and a central server.)
“a merging module implemented on a server distinct from the at least one server of the intermediate processing system;” (FPPDL Framework, pp. 4209; “Each Fog node j first combines all the received data from s connected end nodes, during each epoch of training, it downloads the current model
w
g
l
o
b
a
l
from Cloud, trains independently on the collected data using the standard SGD algorithm to compute model updates
Δ
w
i
, and then sends the selected model updates
Δ
w
i
F
to the Cloud, which sums over all the updates to update
w
g
l
o
b
a
l
, as shown in step 2 conducted by the Cloud.” This model discloses the use of fog nodes to process data and a cloud server which is able to aggregate training information for the node servers. The merging unit is interpreted to be the cloud server distinctly different that the fog nodes.)
“wherein: the at least one digital device is configured to extract blocks of partial data from a first item of content and distribute the blocks of partial data to respective different processing modules in the intermediate processing system;” (FPPDL Framework, pp. 4209; “A descriptive sketch of the training procedure for FPPDL appears in Algorithm 1. As indicated by steps 1–3 conducted by IoT end nodes, all the original data remain at the end nodes, and only the projected data are sent to the nearby Fog node for training.” Algorithm 1 discloses that the end node will partition data into different sets and then send that data to a fog nod for processing. The edge nodes will process the data to ensure that raw data is not sent to the fog node.)
“each of said respective different processing modules receiving a block of partial data from said first item of content is configured to: apply processing to the block of partial data to generate a partial model for an artificial intelligence module, and” (Algorithm 1: FPPDL Training Procedure., pp. 4209; The algorithm lists the method for each fog node in the system. The for not will receive data from different edge nodes and process the data that they sent. The fog nodes are able to train and send updates and weight parameters of models to the cloud server for aggregation.)
Lyu fails to explicitly disclose, “a source of training data, said source of training data comprising at least one digital device;”, “transmit the partial model to the merging module; and” and “the merging module is configured to construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said respective different processing modules.”
However, Ye discloses, “a source of training data, said source of training data comprising at least one digital device;” (A General Framework, pp. 23922; "Vehicular clients are equipped with a set of built-in sensors, such as cameras, GPS, tachographs, lateral acceleration sensors, and also accommodate storage space, computation and communication resources [18]. The built-in sensors are used to capture images that may be preprocessed for data augment." In this article they disclose a system that contains vehicle and a global server. The vehicle contains may sensors both analog and digital. The different sensors are able to send different forms of image data and corresponding information to an onboard processing system.)
“transmit the partial model to the merging module; and” (Selective Model Aggregation, pp. 23924; "To meet synchronization requirements, the updated local DNN models
w
1
E
(
0
)
and
w
2
E
(
0
)
are sent to the central server in time." Each of the vehicles models are transmitted to a central server where they are evaluated and aggregated into a new global model.)
“the merging module is configured to construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said respective different processing modules.” (Selective Model Aggregation, pp. 23924; "After receiving the updated local DNN models
w
1
E
(
0
)
and
w
2
E
0
,
the central server aggregates them to update the global DNN model, which generates the global DNN model w(1)." The central server will take in all partial models from different vehicles and process them. The central server will then aggregate the different models into a global model.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lyu and Ye. Lyu teaches a computing system that utilizes distributed learning that uses fog nodes, which are servers, to take in data from IoT devices and process it and further Lyu contains a central server similar to a federated learning network able to aggregate model parameters from the fog nodes. Ye teaches a federated learning system related to vehicles. Ye teaches a system of sensors which send data to a processing node or server for further processing and then finally the server models are aggregated at a global server. One of ordinary skill would have motivation to combine a fog computing system which is a federated learning system and a federated learning system that uses intermediate servers to process data from sensors, “As shown in Fig. 5, using the MNIST dataset, we compare the accuracy of model aggregation for the CA and FedAvg approaches under BB and BU. As the number of global iteration increases, the accuracy of model aggregation is increasing for the BB and BU cases. The accuracy of model aggregation in the BB case is higher than that in the BU case. In the BB case, because the level of training image quality is closer to the level of testing image quality, which causes a high accuracy in classifying the images. In the BU case, because the gap between the level of training image quality and the level of testing image quality is large, which leads to a low accuracy in classifying the images. The similar results appear in [5]. In the BB case, the accuracy of model aggregation with the CA approach is 2.42% higher than the accuracy of model aggregation with the FedAvg approach. In the BU case, the accuracy of model aggregation adopting the CA approach is 6.28% higher than the accuracy of model aggregation adopting the FedAvg approach.” (Ye, Accuracy and efficiency of model aggregation, pp. 23929-23930).
Regarding Claim 2, Ye discloses, “wherein said at least one digital device is configured to extract blocks of partial data from the first item of content by extraction of specific target features in the item of content.” (a General Framework, pp. 23922; "Vehicular clients are equipped with a set of built-in sensors, such as cameras, GPS, tachographs, lateral acceleration sensors, and also accommodate storage space, computation and communication resources [18]. The built-in sensors are used to capture images that may be preprocessed for data augment." This system uses many different sensors to capture data. Specifically, these sensors are used to capture image data for processing and labeling. This data is evaluated and refined to be used in generating a local model.)
Regarding Claim 3, Ye discloses, “wherein said at least one digital device is configured to extract blocks of partial data from the first item of content by dividing the first item of content into blocks of partial data, the blocks having a specified size.” (Computation Capability, pp. 23925; “For vehicular client m, contributing images and computation resources incurs a cost of resource consumption, which is denoted as:
c
k
,
m
=
α
k
,
m
x
k
,
m
+
E
k
.
m
e
k
,
m
x
k
,
m
f
k
,
m
2
, where
α
k
,
m
is the unit cost for collecting each image,
x
k
,
m
is the amount of images, and
f
k
,
m
is the amount of computation resources.
E
k
is regarded as a constant for all the vehicular clients [14], [24]. According to [25],
e
k
,
m
=
l
k
,
m
b
k
,
m
η
η
k
,
m
ρ
k
,
m
where
l
k
,
m
is the unit cost for the computation resource consumption,
b
k
,
m
is the size of each image,
η
k
,
m
is the effective switched capacitance that depends on the chip architecture, and
ρ
k
,
m
is the number of CPU cycles to process one bit.” The images are evaluated by the processing system. This will take in individual images from a set of images and process them. It is also mentioned that the image data is of a specified size labeled as
b
k
,
m
.)
Regarding Claim 4, Ye discloses, “wherein said at least one digital device is configured to transmit ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said respective different processing modules are configured to transmit to the merging module their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.” (Image Quality, pp. 23924; “According to the model, the motion blur level can be implicitly predicted by observing the instantaneous velocity of each vehicular client. By [7], we have: [see equation (1)], where
v
'
is the relative velocity between velocity v of vehicular client and velocity
v
0
of the object,
σ
is the perpendicular distance from the pinhole to the starting point of an object, l is the length of the motion blur on the image plane, H is the exposure time interval, s is the camera focal length,
δ
is the angle between the image plane and the motion direction, and g is the starting position of the object on the image plane. We denote the charge-coupled device (CCD) pixel size in the horizontal direction as Q, and have: [see equation (2)], where G and L are the starting position of the object and the level of motion blur in the image (in pixels), respectively.” The image data is evaluated and processed by the system. The image data is of moving images and the data may be blurred or unusable. The images are collected while moving at times and the processing system will evaluate many different images in a set of images. The different values of the data, including positional data is kept by the system. This in turn will affect the local model when it is generated from the image data.)
Regarding Claim 8, Lyu discloses, A processing system configured for use in an AI-module training system, said AI-module training system comprising at least one data-source device and a merging module implemented on a server, the processing system comprising:” (FPPDL Framework, pp. 4209; “We display a three-level Fog-embedded framework for deep learning in Fig. 1(d), which can support numerous end nodes at the bottom level, Fog nodes in the middle Fog level, and the Cloud at the top level. The Fog nodes have two functions: (i) as Fog servers with more powerful storage and processing capability than end nodes; (ii) as Fog edge nodes that can interact with heterogenous devices with different protocols [17].” This article discloses a system that trains AI models in a distributed system. This system contains a three-level architecture which is an end node, fog node and a central server.)
“a plurality of processing modules implemented on at least one server distinct from the server implementing the merging module, each of the respective plurality of processing modules configured to receive respective blocks of partial data from said at least one data-source device, said blocks of partial data being extracted from a common item of digital content;” (FPPDL Framework, pp. 4209; “We display a three-level Fog-embedded framework for deep learning in Fig. 1(d), which can support numerous end nodes at the bottom level, Fog nodes in the middle Fog level, and the Cloud at the top level. The Fog nodes have two functions: (i) as Fog servers with more powerful storage and processing capability than end nodes; (ii) as Fog edge nodes that can interact with heterogenous devices with different protocols [17].” This article discloses a three-level system which contains multiple end nodes and user, multiple fog nodes, and a central server able to merge and generate global models.)
Lyu fails to explicitly disclose, “wherein each of the respective different processing modules is configured to: apply processing to its respective received block of partial data to generate a partial model for an artificial intelligence module, and” and “transmit the partial model to the merging module of the AI-module training system.”.
However, Ye discloses, “wherein each of the respective different processing modules is configured to: apply processing to its respective received block of partial data to generate a partial model for an artificial intelligence module, and” (Selective Model Aggregation, pp. 23924; "According to the predesigned contract items, vehicular clients 1 and 2 train a local DNN model by using their local images and computation resources. More specifically, vehicular client 1 uses the global DNN model w(O) and a number of xl local images to conduct the forward-backward propagation algorithm to minimize the local loss function F1(w(O))." When generating a local model the system will collect and consolidate the data taken from the sensors and process it. This processed data will be used to generate a local model, which under the broadest reasonable interpretation would be considered a partial model.)
“transmit the partial model to the merging module of the AI-module training system.” (Selective Model Aggregation, pp. 23924; "After receiving the updated local DNN models
w
1
E
(
0
)
and
w
2
E
0
,
the central server aggregates them to update the global DNN model, which generates the global DNN model w(1)." The central server will take in all partial models from different vehicles and process them. The central server will then aggregate the different models into a global model.)
Regarding Claim 9, Ye discloses, “wherein said processing modules are configured to receive ancillary data in association with the blocks of partial data from the first item of content, said ancillary data comprising positional data indicative of the position of the respective partial block of data within the first item of content, and said processing modules are configured to transmit to the merging module of the AI-module training system their partial models and the positional data associated with the blocks of partial data from which the partial models were produced.” (Image Quality, pp. 23924; “According to the model, the motion blur level can be implicitly predicted by observing the instantaneous velocity of each vehicular client. By [7], we have: [see equation (1)], where
v
'
is the relative velocity between velocity v of vehicular client and velocity
v
0
of the object,
σ
is the perpendicular distance from the pinhole to the starting point of an object, l is the length of the motion blur on the image plane, H is the exposure time interval, s is the camera focal length,
δ
is the angle between the image plane and the motion direction, and g is the starting position of the object on the image plane. We denote the charge-coupled device (CCD) pixel size in the horizontal direction as Q, and have: [see equation (2)], where G and L are the starting position of the object and the level of motion blur in the image (in pixels), respectively.” The image data is evaluated and processed by the system. The image data is of moving images and the data may be blurred or unusable. The images are collected while moving at times and the processing system will evaluate many different images in a set of images. The different values of the data, including positional data is kept by the system. This in turn will affect the local model when it is generated from the image data.)
Regarding Claim 11, Lyu discloses, “A digital device configured for use as a source of training data for an artificial-intelligence-module training system comprising the intermediate processing system of claim 8,” (FPPDL Framework, pp. 4209; “We display a three-level Fog-embedded framework for deep learning in Fig. 1(d), which can support numerous end nodes at the bottom level, Fog nodes in the middle Fog level, and the Cloud at the top level. The Fog nodes have two functions: (i) as Fog servers with more powerful storage and processing capability than end nodes; (ii) as Fog edge nodes that can interact with heterogenous devices with different protocols [17].” This system includes digital devices that are send data to nodes for training. These processing nodes are given a pretrained model initially which was trained using training data.)
“wherein said digital device is configured to: extract blocks of partial data from a first item of content; and” (FPPDL Framework, pp. 4209; “A descriptive sketch of the training procedure for FPPDL appears in Algorithm 1. As indicated by steps 1–3 conducted by IoT end nodes, all the original data remain at the end nodes, and only the projected data are sent to the nearby Fog node for training.” Algorithm 1 discloses that the end node will partition data into different sets and then send that data to a fog nod for processing. The edge nodes will process the data to ensure that raw data is not sent to the fog node.)
Lyu fails to explicitly disclose, “distribute the blocks of partial data to respective different processing modules in the intermediate processing system, so that each of said different processing modules may apply processing to the block of partial data to generate a partial model for the artificial intelligence module and transmit said partial model to a merging module in the artificial-intelligence-module training system.”.
However, Ye discloses, “distribute the blocks of partial data to respective different processing modules in the intermediate processing system, so that each of said different processing modules may apply processing to the block of partial data to generate a partial model for the artificial intelligence module and transmit said partial model to a merging module in the artificial-intelligence-module training system.” (Selective Model Aggregation, pp. 23924; "According to the predesigned contract items, vehicular clients 1 and 2 train a local DNN model by using their local images and computation resources. More specifically, vehicular client 1 uses the global DNN model w(O) and a number of xl local images to conduct the forward-backward propagation algorithm to minimize the local loss function F1(w(O))." When generating a local model the system will collect and consolidate the data taken from the sensors and process it. This processed data will be used to generate a local model, which under the broadest reasonable interpretation would be considered a partial model.)
Regarding Claim 12, Lyu discloses, A merging module implemented on a server and configured for use in an artificial-intelligence-module training system comprising the intermediate processing system of claim 8;” (FPPDL Framework, pp. 4209; “We display a three-level Fog-embedded framework for deep learning in Fig. 1(d), which can support numerous end nodes at the bottom level, Fog nodes in the middle Fog level, and the Cloud at the top level. The Fog nodes have two functions: (i) as Fog servers with more powerful storage and processing capability than end nodes; (ii) as Fog edge nodes that can interact with heterogenous devices with different protocols [17].” This model contains a merging module which is interpreted to be the cloud server. This margining module will take the updates of the models located at the nodes and generate a new global model.)
Lyu fails to explicitly disclose, “wherein said merging module is configured to: receive partial models from different processing modules in the intermediate processing system; and” and “construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said different processing modules.”.
However, Ye discloses, “wherein said merging module is configured to: receive partial models from different processing modules in the intermediate processing system; and” (Selective Model Aggregation, pp. 23924; “After receiving the updated local DNN models
w
1
E
(0) and
w
2
E
(0), the central server aggregates them to update the global DNN model, which generates the global DNN model w(1).” The central server will take in all partial models from different vehicles and process them. The central server will then aggregate the different models into a global model.)
“construct a global model of the artificial intelligence module by performing processing operations including processing the partial models received from said different processing modules.” (Selective Model Aggregation, pp. 23924; “After receiving the updated local DNN models
w
1
E
(0) and
w
2
E
(0), the central server aggregates them to update the global DNN model, which generates the global DNN model w(1).” After the central server has taken in all partial models from different vehicles it will process them. The central server will then aggregate the different models into a global model.)
Regarding Claim 13, Lyu discloses, “A computer-implemented method to train an artificial intelligence module, the method comprising:” (Algorithm 1: FPPDL Training Procedure, pp. 4209; This algorithm discloses a training method for training local and global federated learning models in a fog computing system. This algorithm discloses the process for the client, fog nods and the cloud server.)
“a) extracting blocks of partial data from a first item of content stored or acquired at a digital device,” (FPPDL Framework, pp. 4209; “A descriptive sketch of the training procedure for FPPDL appears in Algorithm 1. As indicated by steps 1–3 conducted by IoT end nodes, all the original data remain at the end nodes, and only the projected data are sent to the nearby Fog node for training.” Algorithm 1 discloses that the end node will partition data into different sets and then send that data to a fog nod for processing. The edge nodes will process the data to ensure that raw data is not sent to the fog node.)
“c) applying processing to the block of partial data, by the respective different processing systems, to generate respective partial models for the artificial intelligence module;” (Algorithm 1: FPPDL Training Procedure., pp. 4209; The algorithm lists the method for each fog node in the system. The for not will receive data from different edge nodes and process the data that they sent. The fog nodes are able to train and send updates and weight parameters of models to the cloud server for aggregation.)
Lyu fails to explicitly disclose, “b) distributing the blocks of partial data from the digital device to respective different processing modules in an intermediate processing system;”, “d) transmitting the partial models from the intermediate processing system to a merging module; and” and “e) constructing a global model of the artificial intelligence module by the performance of processing operations by the merging module, said processing operations including processing the partial models received from the respective different processing modules.”
However, Ye discloses, “b) distributing the blocks of partial data from the digital device to respective different processing modules in an intermediate processing system;” (A General Framework, pp. 23922; After that, the preprocessed images are classified and labeled by automatic labeling technology [6], and are cached in vehicular clients.” Each of the different sensors collect data and then send them to a processing unit. This will label the data and prepare it to be used in generating a local model.)
“d) transmitting the partial models from the intermediate processing system to a merging module; and” (Selective Model Aggregation, pp. 23924; "To meet synchronization requirements, the updated local DNN models
w
1
E
(
0
)
and
w
2
E
(
0
)
are sent to the central server in time." Each of the vehicles models are transmitted to a central server where they are evaluated and aggregated into a new global model.)
“e) constructing a global model of the artificial intelligence module by the performance of processing operations by the merging module, said processing operations including processing the partial models received from the respective different processing modules.” (Selective Model Aggregation, pp. 23924; "After receiving the updated local DNN models
w
1
E
(
0
)
and
w
2
E
0
,
the central server aggregates them to update the global DNN model, which generates the global DNN model w(1)." The central server will take in all partial models from different vehicles and process them. The central server will then aggregate the different models into a global model.)
Regarding Claim 15, Lyu discloses, “A non-transitory computer-readable medium having stored thereon instructions which, when executed by a processor, cause the processor to carry out steps c) and d) of the method according to claim 13.” (FPPDL Framework, pp. 4209; “We display a three-level Fog-embedded framework for deep learning in Fig. 1(d), which can support numerous end nodes at the bottom level, Fog nodes in the middle Fog level, and the Cloud at the top level. The Fog nodes have two functions: (i) as Fog servers with more powerful storage and processing capability than end nodes; (ii) as Fog edge nodes that can interact with heterogenous devices with different protocols [17].” The model in this system is able to perform the steps of claim 13. That is this will take in data and transmit the data to fog servers for processing. Further the updates to the model located at the fog server are sent to a cloud server and a new update global model is generated from the updates from the fog nodes.) And (Performance Evaluation, pp. 4211; “Model Architectures: The implementation architecture is the multilayer perceptron (MLP) as in [4], which is a simple multilayer-perceptron having two hidden layers with 128 and 64 units using ReLU activations. The exact specification of the MLP architecture is detailed in Fig. 2. There are a total of 1 40 106 parameters for MNIST, and 402 250 parameters for SVHN [4].” The method disclosed in this article is executed using generic computing systems. The model is able to use previously known machine learning models and datasets which are implemented on generic computing systems.)
Regarding Claim 16, Lyu discloses, “wherein no processing module of the plurality of processing modules receives the entire first item of content.” (FPPDL Framework, pp. 4210; “Next, we discuss the key components used in FPPDL: end nodes apply Random Projection (RP) to the original data before forwarding them to the nearby Fog node, then Fog nodes apply DPSGD to train differentially private Fog-level models.” This system will mutate and partition the data before it is sent to the node server for further processing. This would teach that the node never receives the entire and complete data item from the end node. The raw data from the user will remain at edge node.)
Regarding Claim 18, Lyu discloses, “wherein the merging module is configured to perform subsequent steps of the convolution neural network algorithm not performed by the plurality of processing modules to construct the global model of the artificial intelligence module.” (Algorithm 1: FPPDL Training procedure, pp. 4209; The model in this article will receive all of the model updates form the intermediate nodes. This will then perform updates to the global model based on the updates and generate a new model. This is seen at lines 1-2 in the Role: cloud section of the algorithm. The final layer of a CNN is usually a fully connected layer, which is a layer containing a neural network containing nodes and connections. The cloud in this model is able to use the data form the nodes to produce an updated neutral network.)
Claims 5-7, and 10 are rejected under 35 U.S.C. 103 as being unpatentable over Ye and Lyu in view of Vepakomma et al, (Vepakomma et al, “Split learning for health: Distributed deep learning without sharing raw patient data”, 2018, hereinafter “Vepakomma”).
Regarding Claim 5, Ye and Lyu fail to explicitly disclose the elements of this claim, however, Vepakomma discloses, “wherein each of said respective different processing modules is configured to apply a convolution operation to its received block of partial data from said first item of content.” (Results about resource efficiency, pp. 3; "In the case with a relatively smaller number of clients the communication bandwidth required by federated learning is less than splitNN. These improvements on the client-side resource efficiency are even more dramatic due to the presence of a smaller number of parameters in earlier layers of convolutional neural networks (CNNs) like VGG and Resnet in addition to the fact that computation is split due to the cut layers." This system can work with Convolutional neural networks. The layers of the CNN can be split across multiple devices as see in figure 2. Under the broadest reasonable interpretation while using a CNN network with this system it would be obvious to perform convolutional operations between the layers.) and (Vertically partitioned data for split learning, pp. 3; "As a concrete example we walkthrough the case where radiology centers collaborate with pathology test centers and a server for disease diagnosis. As shown in Fig. 2c radiology centers holding imaging data modalities train a partial model up to the cut layer. In the same way the pathology test center having patient test results trains a partial model up to its own cut layer." This article discloses different methods of splitting a neural network. Figure 2c shows the different layers of the model and the clients who add their data to the different layers. This teaches that the data received from the clients will be combined to create a partial model.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lyu, Ye and Vepakomma. Lyu teaches a computing system that utilizes distributed learning that uses fog nodes, which are servers, to take in data from IoT devices and process it and further Lyu contains a central server similar to a federated learning network able to aggregate model parameters from the fog nodes. Ye teaches a federated learning system related to vehicles. Ye teaches a system of sensors which send data to a processing node or server for further processing and then finally the server models are aggregated at a global server. Vepakomma teaches a split learning method where a machine learning model is split and processed in different location while preserving the privacy of users. One of ordinary skill would have motivation to combine a fog computing system which is a federated learning system and a federated learning system that uses intermediate servers to process data from sensors, with a system that is able to assist in generate strategies for splitting a neural network processing the data across different devices, “In this distributed learning experiment we clearly see that SplitNN outperforms the techniques of federated learning and large batch synchronous SGD in terms of higher accuracies with drastically lower computational requirements on the side of clients. In tables 1 and 2 we share more comparisons from [32] on computing resources in TFlops and communication bandwidth in GB required by these techniques. SplitNN again has a drastic improvement of computational resource efficiency on the client side. In the case with a relatively smaller number of clients the communication bandwidth required by federated learning is less than splitNN. These improvements on the client-side resource efficiency are even more dramatic due to the presence of a smaller number of parameters in earlier layers of convolutional neural networks (CNNs) like VGG and Resnet in addition to the fact that computation is split due to the cut layers." (Vepakomma, Results about resource efficiency, pp. 3).”
Regarding Claim 6, Ye and Lyu fail to explicitly disclose the elements of this claim, however, Vepakomma discloses, “wherein the merging module is configured to construct a global model of the artificial intelligence module by performing a series of pooling and convolution operations on the partial models received from said respective different processing modules, and by training a fully-connected neural network using the results of said pooling and convolution operations.” (Vertically partitioned data for split learning, pp. 3; "The outputs at the cut layer from both these centers are then concatenated and sent to the disease diagnosis server that trains the rest of the model. This process is continued back and forth to complete the forward and backward propagations in order to train the distributed deep learning model without sharing each other’s raw data. We would like to note that although these example configurations show some versatile applications for splitNN, they are by no means the only possible configurations." Figure 2c discloses a split neural network using vertically partitioned data. The lower section of figure 2c shows the combination of the two client's models into a global model.)
Regarding Claim 7, Ye and Lyu fail to explicitly disclose the elements of this claim, however, Vepakomma discloses, “wherein the intermediate processing system comprises a set of servers implementing said processing modules.” (Simple vanilla configuration for split learning, pp. 2; "This is the simplest of splitNN configurations as shown in Fig 2a. In this setting each client, (for example, radiology center) trains a partial deep network up to a specific layer known as the cut layer. The outputs at the cut layer are sent to a server which completes the rest of the training without looking at raw data (radiology images) from clients." This example used in this article discloses a simple split of a neural network between a client and a server. This example uses health entities as the server which in more complex implementations of this system could contain multiple servers and clients.)
Regarding Claim 10, Ye and Lyu fail to explicitly disclose the elements of this claim, however, Vepakomma discloses, “wherein each of said processing modules is configured to apply a convolution operation to its respective received block of partial data from said first item of content.” (Vertically partitioned data for split learning, pp. 3; "As a concrete example we walkthrough the case where radiology centers collaborate with pathology test centers and a server for disease diagnosis. As shown in Fig. 2c radiology centers holding imaging data modalities train a partial model up to the cut layer. In the same way the pathology test center having patient test results trains a partial model up to its own cut layer." This article discloses different methods of splitting a neural network. Figure 2c shows the different layers of the model and the clients who add their data to the different layers. This teaches that the data received from the clients will be combined to create a partial model.)
Claim 17 is rejected under 35 U.S.C. 103 as being unpatentable over Ye and Lyu in view of Mehta et al, (Mehta et al, “DeepSplit: Dynamic Splitting of Collaborative Edge-Cloud Convolutional Neural Networks”, 2020, hereinafter “Mehta”).
Regarding Claim 17, Ye and Lyu fail to explicitly disclose the elements of this claim, however, Mehta discloses, “wherein each of the plurality of processing modules are configured to perform at least a first convolution operation of a convolution neural network algorithm to generate respective partial models for the artificial intelligence module.” (Edge-Cloud Collaborative, pp. 722, “We achieve CNN splitting and sharing of the partial results with the help of the Neurosurgeon approach introduced by Kang et al. [1]. We first store a fixed number of layers at edge. When the new input image frame arrives, we calculate its partial result with layers present at edge. The partial output is a Numpy array that we store in a file. The file is then sent to cloud. In the cloud, the partial output read from the file is fed in as input to the next CNN layer in the cloud. The final result computed is communicated back to edge for decision making process.” The model in this article is able to split layers of a CNN and perform certain layers on an edge device and the remaining layers on a cloud server. This method uses a CNN model and would perform CNN operation on the edge device and then send the output of the last layer before the split to a could to complete the computation.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lyu, Ye and Mehta. Lyu teaches a computing system that utilizes distributed learning that uses fog nodes, which are servers, to take in data from IoT devices and process it and further Lyu contains a central server similar to a federated learning network able to aggregate model parameters from the fog nodes. Ye teaches a federated learning system related to vehicles. Ye teaches a system of sensors which send data to a processing node or server for further processing and then finally the server models are aggregated at a global server. Mehta teaches a split leaning model which is able to split a CNN operation between an end node and a cloud server. One of ordinary skill would have motivation to combine a fog computing system which is a federated learning system and a federated learning system that uses intermediate servers to process data from sensors, with a system that is able to split computation of a CNN across different processing devices in a network, “Figure 5 shows how higher input image dimensions can lead to higher bandwidth usage. Figure 7 emphasizes the side effects of CNN splitting, as it introduces large layer weight transfer overheads between edge and cloud. When we perform splitting, we need to move entire layers between edge and cloud. Moving layers is equivalent to moving layer weights. Figure 7 shows such an example layer transfers. We start with 27 YOLO V2 CNN layers at edge and one remaining layer in the cloud. We then sequentially transfer 26 layers from edge to cloud and at the end are left with one layer at edge and 27 layers in the cloud. The graph shows the amount of data overhead we introduce when we perform such layer transfers from edge to cloud. The key takeaway here is that moving many layers between edge and cloud simultaneously can introduce large overheads, and hence should be avoided. Also, splitting should be performed infrequently, only in cases when it’s absolutely necessary, such as when a new critical task arrives and we need to preempt a fraction of the memory. Dynamic CNN splitting is very effective when done infrequently, hence we suggest setting load thresholds upon crossing which dynamic splitting would take place.” (Mehta, Performance Evaluations, pp. 725)
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PAUL MICHAEL GALVIN-SIEBENALER whose telephone number is (571)272-1257. The examiner can normally be reached Monday - Friday 8AM to 5PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Viker Lamardo can be reached at (571) 270-5871. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/PAUL M GALVIN-SIEBENALER/Examiner, Art Unit 2147
/VIKER A LAMARDO/Supervisory Patent Examiner, Art Unit 2147
Prosecution Insights