DETAILED ACTION
1. This communication is in response to the amendments filed on March 16, 2026 for Application No. 18/022,216 in which Claims 1, 17-24, and 26-33 are presented for examination.
Notice of Pre-AIA or AIA Status
2. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
3. The amendments filed on March 16, 2026 have been considered. Claims 1, 18, 27-28, and 33 have been amended. Claim 25 has been cancelled. Thus, Claims 1, 17-24 and 26-33 are pending and presented for examination.
4. Applicant's arguments filed March 16, 2026 with respect to the 35 U.S.C. 101 rejection have been fully considered but they are not persuasive.
Applicant’s Arguments on Pgs. 2-3 of Arguments/Remarks state:
“Page 2 of the Eligibility memo starts with how "Examiners should determine whether a claim satisfies the criteria for subject matter eligibility by evaluating the claim in accordance with the flowchart provided in MPEP 2106," which includes evaluations under Step 1, Step 2A (Prong One and Prong Two), and Step 2B.
Step 2A of the 2019 Revised Patent Subject Matter Eligibility Guidance is a two-prong inquiry. In Step 2A Prong One, the Office evaluates whether the claim recites a judicial exception, such as an "abstract idea," as alleged on page 3 of the Office Action. Specifically, the Office Action rejects independent claims 1 and 18 because the features of the claims are directed to "mental processes." (See, pages 3-6 of the Office Action). However, the Eligibility memo notes "that a claim recites a mental process when it contains limitation(s) that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions" and that "a claim does not recite a mental process when it contains limitation(s) that cannot practically be performed in the human mind, for instance when the human mind is not equipped to perform the claim limitation(s)." (See, page 2 of the Eligibility memo). The Eligibility memory further emphasizes that "The mental process grouping is not without limits. Examiners are reminded not to expand this grouping in a manner that encompasses claim limitations that cannot practically be performed in the human mind." Id. Following the logic outlined in the Eligibility memo, independent claim 1 is not directed to an abstract idea and cannot practically be performed in the human mind. Instead, the claims recite a specific methods and devices for "transferring] learning from two or more source domains including a first source domain and a second source domain". For example, claim 1 recites "generating a first data by using a first decoder model with a first set of target features, wherein the first decoder model is based on the first source domain," "updating a final set of target features and final data based on the generated first data," "generating a second data by using a second decoder model with a second set of target features, wherein the second data that is generated is conditioned on the first set of target features and wherein the second decoder model has been trained by the second source domain conditionally on a subset of features common to the second set of target features and the first set of target features," "updating the final set of target features and final data based on the generated second data," as well as "training a target-domain model using the final data and the final set of target features." One of ordinary skill in the art would understand that such a combination steps are not a process that can be practically performed in the human mind. Instead, one of ordinary skill in the art would appreciate that machine learning requires processing data at a scale that is not reasonably possible by the human mind. Thus, one of ordinary skill would understand that the steps of independent claims 1 and 18 are not that in which the human mind is not equipped to perform and thus is not directed to an abstract idea.”
Examiner respectfully disagrees. At Step 2A Prong 1, the limitations “generating a first data […] with a first set of target features […]”, “updating a final set of target features and final data based on the generated first data”, “generating a second data […] with a second set of target features […]”, and “updating the final set of target features and final data based on the generated second data” are recited at a high-level of generality and may be practically performed by mental process. For example, a user may observe/analyze features of a first/second source domain and accordingly use judgement/evaluation to generate first/second data with a first/second set of target features based on said analysis. Further, a user is capable of updating a final set of target features and final data based on generated first/second data by observing/analyzing generated first/second data and accordingly using judgement/evaluation to update a final set of target features and final data (with the aid of pen and paper) based on said analysis of the generated first/second data. Further examples of how these limitations may be feasibly performed by mental process are provided in the 35 U.S.C. 101 section below. Thus, the aforementioned limitations may be practically performed by mental process.
Furthermore, the mere inclusion of generic computer components (i.e., first decoder model, second decoder model, target-domain model, etc.) does not preclude such a mental process interpretation of the aforementioned limitations. Applicant states that “One of ordinary skill in the art would understand that such a combination of steps are not a process that can be practically performed in the human mind. Instead, one of ordinary skill in the art would appreciate that machine learning requires processing data at a scale that is not reasonably possible by the human mind” – but again, the mere inclusion of generic machine learning models does not preclude the limitations of the Independent claims from being performed by mental process. The claims simply state “generating a first data […]”, “updating a final set of target features and final data based on the generated first data”, “generating a second data […]”, and “updating the final set of target features and final data based on the generated second data” – nowhere in the currently drafted claim language do these limitations “require processing data at a scale that is not reasonably possible by the human mind”, instead the claims are drafted at a high-level of generality and merely include generic computer components to perform these operations without significantly more, such that they may be interpreted as mere mental process.
Examiner additionally notes that at Step 2A Prong 2 and Step 2B, the claims recite “[…] by using a first decoder model […] wherein the first decoder model is based on the first source domain”, “by using a second decoder model […] wherein the second decoder model has been trained by the second source domain conditionally on a subset of features common to the second set of target features and the first set of target features”, and “training a target-domain model using the final data and the final set of target features”. These limitations all amount to merely adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f). The first decoder model is simply “based on the first source domain” without significantly more, the second decoder model is already trained/configured generically and “conditionally on a subset of features common […]”, and the target-domain model is merely trained “using the final data and the final set of target features” without significantly more. The architecture/composition/training process of the models are not further detailed by the claims – instead, the claims generically describe how the models are “trained” simply “using” previously determined data without significantly more. These models may just amount to off-the-shelf black box models which Applicant merely “uses” to generate first/second data and correspondingly update features and data. This cannot provide an inventive concept.
Applicant’s Arguments on Pgs. 3-4 of Arguments/Remarks state:
“Even if, arguendo, the claim were to be considered directed to an abstract idea under Step 2A, Prong One, independent claims 1 and 18 satisfy Step 2A, Prong Two because the alleged abstract idea is integrated into a practical application. The Eligibility memo notes that, in Step 2A Prong Two, the Office should consider "the claim as a whole" such that "the additional limitations should not be evaluated in a vacuum, completely separate from the recited judicial exception. Instead, the analysis should take into consideration all the claim limitations and how these limitations interact and impact each other when evaluating whether the exception is integrated into a practical application." (See, pages 2-3 of the Eligibility memo). The Eligibility memo further notes how the Office "can conclude that claims are eligible in Step 2A Prong Two by finding that a claim reflects an improvement to the functioning of a computer or to another technology or technical field, integrating a recited judicial exception into a practical application of the exception." Lastly, in the Eligibility memo, the "examiner is reminded to consult the specification to determine whether the disclosed invention improves technology or a technical field, and evaluate the claim to ensure it reflects the disclosed improvement" noting that "The claim itself does not need to explicitly recite the improvement described in the specification."
Paragraphs [0005]-[0011] of Applicant's Specification discuss issues that may be present when performing transfer learning, particularly, as it relates to inter-radio access technologies, and identifies how the claimed invention can be implemented to provide an improvement to how existing knowledge can be used when transferring a model between similar types of systems based on different underlying technologies (e.g., 3G, 4G, 5G, etc.). Therefore, independent claims 1 and 18 reflect an improvement to the functioning of a computer and/or to another technology or technical field, integrating any alleged judicial exception into a practical application of that exception.
Moreover, even if, arguendo, the additional elements do not integrate the exception into a practical application, then the Office must evaluate whether the claim provides an inventive concept, under Step 2B analysis. As noted in the Eligibility memo, "the examiner should consider whether the technological limitations are being used as a tool to improve the recited judicial exception ... or whether the claim as a whole provides an improvement to technology or a technical field. Claims that are determined to improve computer capabilities or improve technology or a technical field support a finding that the claim integrates the judicial exception into a practical application or amounts to significantly more than the judicial exception itself." Independent claims 1 and 18 provide a non-conventional arrangement of steps to "transfer learning from two or more source domains including a first source domain and a second source domain." This is not a routine or conventional use of a generic computer, but rather a specific logic flow that provides systems and methods "for seamless handover of existing technologies (e.g., 2G, 4G) to newer technologies (e.g., 4G, 5G, and beyond.)". Therefore, the combination of steps for Independent claims 1 and 18 to "transfer learning from two or more source domains including a first source domain and a second source domain" satisfies the "significantly more" requirement.
For at least the reasons discussed above, Applicant respectfully submits that independent claims 1 and 18 patent-eligible. Accordingly, Applicant respectfully requests the rejection of Claims 1 and 18, and the claims depending therefrom, be withdrawn and the claims allowed.”
Examiner respectfully disagrees for substantially the same reasons as stated above. Regarding Applicant’s citation to a supposed technological improvement, although the claims themselves do not need to explicitly recite the improvement described in the specification, the claims still need to reflect the improvement – the instant claims do not reflect any such improvement. As outlined in the preceding response to arguments above, Applicant’s currently drafted claim language is very generic/high-level and simply amounts to steps of generating and updating generic data using generic/black box machine learning models without significantly more. As explained above, this cannot provide an inventive concept. Moreover, such an improvement to “transfer learning” for different underlying technologies is not recognized/reflected by the currently drafted claim language – the claim just seems to recite standard/generic “transfer learning”. Applicant is encouraged to further amend the claims to better describe the training/configuration of the machine learning models, such that the improvement to transfer learning is apparent in the drafted claim language.
Thus, the 35 U.S.C. 101 rejection is maintained.
5. Applicant's arguments filed March 16, 2026 with respect to the 35 U.S.C. 103 rejection have been fully considered but they are not persuasive.
Applicant’s Arguments on Pgs. 5-6 of Arguments/Remarks state:
“Applicant submits that the combination of Jawahar and Lee fail to teach or suggest at least the above-noted features of independent claims 1 and 18. With respect to the previously presented claim 25, the Office alleged that paragraph [0099] of Jawahar disclosed these features. Applicant respectfully disagrees. Jawahar discusses how, "the adaptation processor 208 may be configured to formulate a second neural network based on the pseudo-labeled instances of the target domain and the learned common representation. The pseudo-labeled instances of the target domain may constitute a second input layer of the second neural network and the learned common representation may constitute a second hidden layer of the second neural network. Further, the determination of the target specific representation by the adaptation processor 208 may correspond to an iterative process, such that the target specific representation may be updated in each iteration of the iterative process." (See, paragraph [0099] of Jawahar).
However, the second neural network formulated by the adaptation processor in Jawahar does not involve a "second decoder model [that] has been trained by the second source domain conditionally on a subset of features common to the second set of target features and the first set of target features." Instead, Jawahar discusses how a second neural network is formulated based on a "learned common representation." A learned common representation does not disclose the second neural network being "trained...conditionally on a subset of features," much less that it is trained "conditionally on a subset of features common to the second set of target features and the first set of target features." In fact, as discussed in paragraphs [0118]-[0121] of Jawahar, the common representation relates to the classification of a plurality of unlabeled text segments based on a plurality of labeled text segments. Therefore, Jawahar does not teach or suggest the usage of "a first decoder model with a first set of target features" and "a second decoder model with a second set of target features" being used to train a second decoder model. As such, Jawahar cannot reasonably be construed to teach or suggest the above-noted features of claim 1.
For at least the reasons discussed above, Applicant respectfully submits that independent claims 1 and 18 patentable over the cited art. Accordingly, Applicant respectfully requests the rejection of Claims 1 and 18, and the claims depending therefrom, be withdrawn and the claims allowed.”
Examiner respectfully disagrees. The prior art references of record, Jawahar in view of Less, still teaches the newly added limitation “[…] wherein the second decoder model has been trained by the second source domain conditionally on a subset of features common to the second set of target features and the first set of target features”. First, Jawahar Par. [0099] states “In an embodiment, the adaptation processor 208 may be configured to formulate a second neural network based on the pseudo-labeled instances of the target domain and the learned common representation. The pseudo-labeled instances of the target domain may constitute a second input layer of the second neural network and the learned common representation may constitute a second hidden layer of the second neural network.” – this is similarly supported by Jawahar Figure 4 which depicts an exemplary scenario on learning transferable feature representations from a plurality of source domains for a target domain. Thus, Jawahar teaches a second model which is trained and configured based on a learned common representation. Furthermore, the learned common representation is described as “[…] a common representation shared between the source domain and the target domain based on the plurality of labeled instances of the source domain […] The common representation may include a plurality of common features extracted from the plurality of labeled instances (such as a text segment, an audio segment, and/or the like) of the source domain that are used in the source domain as well as in the target domain” and this is further supported by Figure 3. Hence, this indicates that the second neural network of Jawahar is trained/configured (see figure 3) by the second source domain conditionally on a subset of features common to the first and second set of target features. While Jawahar does not explicitly disclose the second neural network being a “decoder”, Lee is relied upon for teaching the second neural network comprising a decoder (See Lee Claim 1). Thus, Jawahar in view of Lee teaches the limitations of the instant claims.
Thus, the 35 U.S.C. 103 rejection is maintained.
Claim Rejections - 35 USC § 101
6. 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.
7. Claims 1, 17-24, and 26-33 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1:
Step 1: Claim 1 is a method type claim. Therefore, Claims 1 and 17 are directed to either a process, machine, manufacture, or composition of matter.
2A Prong 1: If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation by mathematical calculation but for the recitation of generic computer components, then it falls within the “Mathematical Concepts” grouping of abstract ideas.
generating a first data […] with a first set of target features […] (mental process – other than reciting “by using a first decoder model”, generating first data may be performed manually by a user observing/analyzing features of a first source domain and accordingly using judgement/evaluation to generate first data based on said analysis. For example, a user may observe/analyze a set of source features associated with an image of a dog (i.e., four legs, fur, tail, etc.) and use judgement/evaluation to generate first data by classifying the image as a dog and/or generating a label that identifies the image as a dog)
updating a final set of target features and final data based on the generated first data (mental process – updating a final set of target features and final data may be performed manually by a user observing/analyzing the generated first data and accordingly using judgement/evaluation to update a final set of target features and final data (with aid of pen and paper) based on said analysis)
generating a second data […] with a second set of target features, wherein the second data that is generated is conditioned on the first set of target features […] (mental process – other than reciting “by using a second decoder model”, generating second data may be performed manually by a user observing/analyzing a second set of target features and accordingly using judgement/evaluation to generate second data based on said analysis. For example, a user may observe/analyze a set of second target features associated with an image of a fish (i.e., fins, gills, etc.) and use judgement/evaluation to generate second data by classifying the image as a fish and/or generating a label that identifies the image as a fish. Furthermore, the second data that is generated may be “conditioned” on the first set of target features (in the aforementioned example, features of a dog), such that the features that are common/uncommon between the first and second set are compared and used to generate the second data based on said analysis)
updating the final set of target features and final data based on the generated second data (mental process – updating the final set of target features and final data may be performed manually by a user observing/analyzing the generated second data and accordingly using judgement/evaluation to update a final set of target features and final data (with aid of pen and paper) based on said analysis)
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
a method for transfer learning from two or more source domains including a first source domain and a second source domain (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: high level recitation of applying transfer learning for a plurality of domains without significantly more)
[…] by using a first decoder model […] wherein the first decoder model is based on the first source domain (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: high level recitation of applying an already trained/configured first decoder model – the first decoder model is merely “based on” the first source domain without significantly more)
[…] by using a second decoder model […] and wherein the second decoder model has been trained by the second source domain conditionally on a subset of features common to the second set of target features and the first set of target features (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: high level recitation of applying an already trained/configured second decoder model – the second decoder model is generically “trained” on previously determined data without significantly more)
training a target-domain model using the final data and the final set of target features (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: high level recitation of training a machine learning model with previously determined data without significantly more)
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
a method for transfer learning from two or more source domains including a first source domain and a second source domain (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: high level recitation of applying transfer learning for a plurality of domains without significantly more. This does not provide an inventive concept)
[…] by using a first decoder model […] wherein the first decoder model is based on the first source domain (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: high level recitation of applying an already trained/configured first decoder model – the first decoder model is merely “based on” the first source domain without significantly more. This does not provide an inventive concept)
[…] by using a second decoder model […] and wherein the second decoder model has been trained by the second source domain conditionally on a subset of features common to the second set of target features and the first set of target features (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: high level recitation of applying an already trained/configured second decoder model – the second decoder model is generically “trained” on previously determined data without significantly more)
training a target-domain model using the final data and the final set of target features (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: high level recitation of training a machine learning model with previously determined data without significantly more. This does not provide an inventive concept)
For the reasons above, Claim 1 is rejected as being directed to an abstract idea without significantly more. This rejection applies equally to dependent claim 17. The additional limitations of the dependent claims are addressed below.
Regarding Claim 17:
Step 2A Prong 1:
See the rejection of Claim 1 above, which Claim 17 depends on.
Step 2A Prong 2 & Step 2B:
[…] enabling transfer learning from two or more source domains according to claim 1 (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner' s note: high level recitation of applying/enabling transfer learning for a plurality of domains without significantly more. This does not provide an inventive concept)
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 1.
Independent Claim 18 recites substantially the same limitations as Claim 1, in the form of a target node, including generic computer components. The claim is also directed to performing mental processes without significantly more, therefore it is rejected under the same rationale.
For the reasons above, Claim 18 is rejected as being directed to an abstract idea without significantly more. This rejection applies equally to dependent claims 19-24 and 26-33. The additional limitations of the dependent claims are addressed below.
Regarding Claim 19:
Step 2A Prong 1:
See the rejection of Claim 18 above, which Claim 19 depends on.
Step 2A Prong 2 & Step 2B:
obtaining a first list of features used by the first source domain (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
obtaining a second list of features used by the second source domain (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
wherein the first set of target features comprises the first list of features and the second set of target features comprises the second list of features (Field of Use – limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application; in this case specifying that the first/second set of target features comprises the first/second list of features does not integrate the exception into a practical application nor amount to significantly more – See MPEP 2106.05(h))
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 20:
Step 2A Prong 1:
See the rejection of Claim 19 above, which Claim 20 depends on.
Step 2A Prong 2 & Step 2B:
sending to a first source domain a first feature list request (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
receiving, in response to the first feature list request, a first list of features used by the first source domain (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 21:
Step 2A Prong 1:
See the rejection of Claim 19 above, which Claim 21 depends on.
Step 2A Prong 2 & Step 2B:
sending to a second source domain a second feature list request (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
receiving, in response to the second feature list request, a second list of features used by second source domain (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 22:
Step 2A Prong 1:
See the rejection of Claim 18 above, which Claim 22 depends on.
Step 2A Prong 2 & Step 2B:
obtaining the first decoder model (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 23:
Step 2A Prong 1:
See the rejection of Claim 22 above, which Claim 23 depends on.
Step 2A Prong 2 & Step 2B:
requesting the first decoder model from the first source domain; and receiving the first decoder model (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 24:
Step 2A Prong 1:
See the rejection of Claim 18, which Claim 24 depends on.
Step 2A Prong 2 & Step 2B:
obtaining the second decoder model (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
wherein the second decoder model has been trained by the second source domain conditionally on the second set of target features (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner' s note: high level recitation of training a machine learning model with previously determined data without significantly more)
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 26:
Step 2A Prong 1:
See the rejection of Claim 24 above, which Claim 26 depends on.
Step 2A Prong 2 & Step 2B:
requesting the second decoder model from the second source domain; and receiving the second decoder model (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 27:
Step 2A Prong 1: See the rejection of Claim 18 above, which Claim 27 depends on.
determining a decoder order sequence based on a number of features that are common among two or more source domains, wherein the decoder order sequence indicates an order in which to generate the first data and the second data (mental process – determining a decoder order sequence may be performed manually by a user observing/analyzing a number of features in order to determine common features between source domains and accordingly using judgement/evaluation to determine an order sequence, in which to generate the first and second data, based on the number of common features between two or more source domains)
Step 2A Prong 2 & Step 2B:
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 28:
Step 2A Prong 1: See the rejection of Claim 19 above, which Claim 28 depends on.
determining a number of features that are common among two or more source domains based on the first list of features and the second list of features (mental process – determining a number of features that are common among two or more source domains may be performed manually by a user observing/analyzing the first and second list of features and accordingly using judgement/evaluation to determine which features are common between the first and second list of features)
determining a decoder order sequence based on the number of features that are common among the two or more source domains, wherein the decoder order sequence indicates an order in which to generate the first data and the second data (mental process – determining a decoder order sequence may be performed manually by a user observing/analyzing a number of features in order to determine common features between source domains and accordingly using judgement/evaluation to determine an order sequence, in which to generate the first and second data, based on the number of common features between two or more source domains)
Step 2A Prong 2 & Step 2B:
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 29:
Step 2A Prong 1:
See the rejection of Claim 18 above, which Claim 29 depends on.
Step 2A Prong 2 & Step 2B:
wherein one or more of the first decoder model and the second decoder model are one of a conditional Generative Adversarial Network (GAN) type model and a conditional Variational Autoencoder (VAE) type model (Field of Use – limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application; in this case specifying that the first/second decoder model are one of a conditional GAN and a conditional VAE does not integrate the exception into a practical application nor amount to significantly more – See MPEP 2106.05(h))
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 30:
Step 2A Prong 1: See the rejection of Claim 18 above, which Claim 30 depends on.
wherein generating a first data by using the first decoder model with the first set of target features comprises filtering data generated by the first decoder model based on a similarity between source and target features (mental process – other than reciting “by using the first decoder model”, generating first data by filtering data may be performed manually by a user observing/analyzing the source and target features and accordingly using judgement/evaluation to determine similarities between source and target features)
wherein generating a second data by using the second decoder model with the second set of target features comprises filtering data generated by the second decoder model based on the similarity between source and target features (mental process – other than reciting “by using the second decoder model”, generating second data by filtering data may be performed manually by a user observing/analyzing the source and target features and accordingly using judgement/evaluation to determine similarities between source and target features)
Step 2A Prong 2 & Step 2B:
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 31:
Step 2A Prong 1:
See the rejection of Claim 30 above, which Claim 31 depends on.
Step 2A Prong 2 & Step 2B:
wherein similarity between source and target features is determined based on one or more distance measures (Field of Use – limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application; in this case specifying that the similarity between the source and target features is based on one or more distance measures does not integrate the exception into a practical application nor amount to significantly more – See MPEP 2106.05(h))
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 32:
Step 2A Prong 1:
See the rejection of Claim 31 above, which Claim 32 depends on.
Step 2A Prong 2 & Step 2B:
wherein the one or more distance measures are selected from the group consisting of a cosine similarity measure, a K-L divergence measure, a Euclidean measure, a Wasserstein measure, and a dot-product measure (Field of Use – limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception does not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application; in this case specifying the types of distance measures does not integrate the exception into a practical application nor amount to significantly more – See MPEP 2106.05(h))
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Regarding Claim 33:
Step 2A Prong 1: See the rejection of Claim 18 above, which Claim 33 depends on.
generating a third data […] with the third set of target features (mental process – other than reciting “by using a third decoder model”, generating third data may be performed manually by a user observing/analyzing the third set of target features and accordingly using judgement/evaluation to generate third data based on said analysis of the third set of target features)
updating the final set of target features and final data based on the generated third data (mental process – updating the final set of target features and final data may be performed manually by a user observing/analyzing the generated third data and accordingly using judgement/evaluation to update a final set of target features and final data (with aid of pen and paper) based on said analysis)
Step 2A Prong 2 & Step 2B:
sending to a third source domain a third feature list request (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
receiving, in response to the third feature list request, a third list of features used by the third source domain (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
requesting a third decoder model with a third set of target features from the third source domain, wherein the third set of target features comprises the third list of features (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
receiving the third decoder model (MPEP 2106.05(d)(II) indicates that merely “Receiving or transmitting data over a network” is a well-understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim). Thereby, a conclusion that the claimed limitation is well-understood, routine, conventional activity is supported under Berkheimer)
wherein the third decoder model has been trained by the third source domain conditionally on the subset of features common to the third set of target features and both the first and second sets of target features (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner's note: high level recitation of applying an already trained/configured third decoder model – the third decoder model is merely “based on” the third source domain without significantly more. This does not provide an inventive concept)
[…] by using the third decoder model […] (mere instructions to apply the exception using generic computer components cannot provide an inventive concept)
Accordingly, under Step 2A Prong 2 and Step 2B, these additional elements do not integrate the abstract idea into practical application because they do not impose any meaningful limits on practicing the abstract idea, as discussed above in the rejection of claim 18.
Claim Rejections - 35 USC § 103
8. 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.
9. Claims 1, 17-24, 26-28, and 30-33 are rejected under 35 U.S.C. 103 as being unpatentable over Jawahar et al. (hereinafter Jawahar) (US PG-PUB 20180218284), in view of Lee et al. (hereinafter Lee) (US PG-PUB 20190114545).
Regarding Claim 1, Jawahar teaches a method for transfer learning from two or more source domains including a first source domain and a second source domain (Jawahar, Par. [0115], “FIG. 4 is a block diagram that illustrates an exemplary scenario for text classification based on learning of transferable feature representations from a plurality of source domains for a target domain, in accordance with at least one embodiment.”, thus, methods (See Jawahar Claim 1 for explicit recitation of a ‘method’) for transfer learning from a plurality of source domains is disclosed), the method comprising:
generating a first data by using a first decoder (Jawahar discloses the model being a neural network but does not explicitly disclose the model being a decoder – See introduction of Lee reference below for explicit recitation of a first decoder model) model with a first set of target features, wherein the first decoder model is based on the first source domain (Jawahar, Par. [0125], “In an embodiment, the data processing server 104 may further use the formulated first neural network 402 for the classification of unlabeled text segments of the source domain. The data processing server 104 may further update the first hidden layer 406 based on the text segments of the source domain that are labeled by the use of the formulated first neural network 402. The result of the classification in the source domain may be stored in the first output layer 408 (“CLASSIFICATION OUTPUT IN SOURCE DOMAIN”).”, thus, first data (result of the classification) is generated by using a first neural network model with a first set of target features (See Figure 4 labels 406A & 406B depicting a source specific representation and common representation comprising target features of the input layer label 404). Further, the first neural network is based on a first source domain, as also supported by Par. [0115] which mentions that the transfer learning is based on a plurality of source domains);
updating a final set of target features and final data based on the generated first data (Jawahar, Par. [0045], “The determination of the target specific representation may correspond to an iterative process, such that the target specific representation may be updated in each iteration of the iterative process. The update of the target specific representation may include addition of new target specific features to the target specific representation in each iteration.” & Par. [0125], “The data processing server 104 may further update the first hidden layer 406 based on the text segments of the source domain that are labeled by the use of the formulated first neural network 402. The result of the classification in the source domain may be stored in the first output layer 408 (“CLASSIFICATION OUTPUT IN SOURCE DOMAIN”)”, therefore, a set of target features is updated based on the generated data at each iteration. Further, as the first neural network is iteratively updated to improve performance, a set of final output data (result of classification) is updated as a result);
generating a second data by using a second decoder (Jawahar discloses the model being a neural network but does not explicitly disclose the model being a decoder – See introduction of Lee reference below for explicit recitation of a second decoder model) model with a second set of target features, wherein the second data that is generated is conditioned on the first set of target features and wherein the second decoder model has been trained by the second source domain conditionally on a subset of features common to the second set of target features and the first set of target features (Jawahar, Par. [0099], ““In an embodiment, the adaptation processor 208 may be configured to formulate a second neural network based on the pseudo-labeled instances of the target domain and the learned common representation. The pseudo-labeled instances of the target domain may constitute a second input layer of the second neural network and the learned common representation may constitute a second hidden layer of the second neural network.” & Par. [0123], “The data processing server 104 may determine the second input layer 414 of the second neural network 412 based on the pseudo-labeled instances of the target domain. Thereafter, the data processing server 104 may extract a plurality of target specific features (e.g., a P-dimensional feature vector of unigrams and bigrams) from the pseudo-labeled instances of the target domain to determine a target specific representation 416A. The common representation 406B and the determined target specific representation 416A may constitute the second hidden layer 416 of the second neural network 412.”, thus, second data (result of the classification) is generated by using a second neural network model with a second set of target features (See Figure 4 label 412 depicting the second neural network comprising a second set of target features label 416A which are used to produce second data/classification result label 418). Further, the second data that is generated is also based on a common representation (See Figure 4 label 406B) such that the second data is conditioned on the first set of target features. Further, the second neural network is based on a second source domain, as also supported by Par. [0115] which mentions that the transfer learning is based on a plurality of source domains);
updating the final set of target features and final data based on the generated second data (Jawahar, Par. [0045], “The determination of the target specific representation may correspond to an iterative process, such that the target specific representation may be updated in each iteration of the iterative process. The update of the target specific representation may include addition of new target specific features to the target specific representation in each iteration.” & Par. [0108], “Thereafter, the adaptation processor 208 may be further configured to update the one or more parameters of the second neural network based on the update in the second hidden layer.”, therefore, a set of target features is updated based on the generated data at each iteration. Further, as the second neural network is iteratively updated to improve performance, a set of final output data (result of classification) is updated as a result); and
training a target-domain model using the final data and the final set of target features (Jawahar, Par. [0046], “In an embodiment, the data processing server 104 may be configured to train a target specific classifier based on the determined target specific representation and the learned common representation. The data processing server 104 may train the target specific classifier after the determination of the target specific representation. The data processing server 104 may further use the trained target specific classifier to perform automatic classification (such as text classification, audio classification, and/or image classification) on the remaining one or more unlabeled instances of the plurality of unlabeled instances of the target domain.”, thus, a target-domain model (target specific classifier) is trained using the final data and final set of target features).
While Jawahar discloses a first neural network and second neural network, Jawahar does not explicitly disclose a first decoder and a second decoder.
However, Lee teaches a first decoder (Lee, Claim 1, “generating a first neural network translation model which includes a neural network having an encoder-decoder structure and learns a feature of source domain data used in an unspecific field;”, thus, a first neural network comprising a first decoder is disclosed) and a second decoder (Lee, Claim 1, “generating a second neural network translation model which includes a neural network having the encoder-decoder structure and learns a feature of target domain data used in a specific field;”, thus, a second neural network comprising a second decoder is disclosed).
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method for transfer learning, as disclosed by Jawahar to include wherein the first and second models comprise a first decoder and a second decoder, as disclosed by Lee. One of ordinary skill in the art would have been motivated to make this modification to enable sequential data processing, through the use of decoders, which may more efficiently decode common features/representations between source and target domains, as compared to traditional neural networks (Lee, Par. [0077], “The decoder 256 may sequentially decode the common feature vector values in the order, in which words constituting the common feature are sorted, to output a common feature-based output vector value corresponding to a translation result of the common feature. The decoder 256 may be configured with a recurrent neural network or a convolutional neural network.”).
Regarding Claim 17, Jawahar in view of Lee teaches a computer-implemented method of enabling transfer learning from two or more source domains (Jawahar, Par. [0115], “FIG. 4 is a block diagram that illustrates an exemplary scenario for text classification based on learning of transferable feature representations from a plurality of source domains for a target domain, in accordance with at least one embodiment.”, therefore, methods for enabling transfer learning from two or more source domains is disclosed) according to claim 1 (See the rejection of Claim 1 above).
Regarding Claim 18, Jawahar in view of Lee teaches a target node, the target node comprising processing circuitry and a memory containing instructions executable by the processing circuitry (Jawahar, Par. [0008], “According to embodiments illustrated herein, there is provided a computer program product for use with a computing device. The computer program product comprises a non-transitory computer readable medium storing a computer program code for domain adaptation for learning transferable feature representations from a source domain for a target domain. The computer program code is executable by one or more processors to receive real-time input data from a computing device over a communication network.”, therefore a target node/computing device (See Applicant’s specification Par. [0029] which uses the term “target node” and “computing device” interchangeably) comprising processing circuitry and a memory containing executable instructions is disclosed)), whereby the processing circuitry is operable to: […]
The rest of the claim language in Claim 18 recites substantially the same limitations as Claim 1, in the form of a target node, therefore it is rejected under the same rationale.
The reasons of obviousness have been noted in the rejection of Claim 1 above and applicable herein.
Regarding Claim 19, Jawahar in view of Lee teaches the target node of claim 18, whereby the processing circuitry is further operable to:
obtaining a first list of features used by the first source domain; and obtaining a second list of features used by the second source domain, wherein the first set of target features comprises the first list of features and the second set of target features comprises the second list of features (Jawahar, Par. [0043], “The data processing server 104 may be further configured to learn a source specific representation corresponding to the source domain and a common representation shared between the source domain and the target domain based on the plurality of labeled instances of the source domain. The source specific representation may include a plurality of source specific features extracted from the plurality of labeled instances (such as a text segment, an audio segment, and/or the like) of the source domain that are used specifically in accordance with the source domain. Examples of the plurality of source specific features may include unigrams, bigrams, special graphical characters, and/or the like.”, therefore, a first/second list of features (See Par. [0024] which mentions that the representation may comprise a feature vector) used by the respective source domains may be obtained via a feature extraction process. Further, as mentioned by Par. [0115], the feature representations may be learned from a plurality of source domains – hence, the features may be obtained for a first/second source domain).
Regarding Claim 20, Jawahar in view of Lee teaches the target node of claim 19, wherein obtaining a first list of features used by the first source domain comprises:
sending to a first source domain a first feature list request; and receiving, in response to the first feature list request, a first list of features used by the first source domain (Jawahar, Par. [0041], “In an embodiment, the data processing server 104 may be configured to receive the classification request from the user-computing device 102 for classification of the plurality of unlabeled instances of the target domain. In an embodiment, the classification request may include the plurality of unlabeled instances of the target domain and the plurality of labeled instances of the source domain.”, therefore, a request may be sent to a first source domain (user inquiring classification regarding a first source domain) and in response, a first list of features/labeled instances of the source domain may be received).
Regarding Claim 21, Jawahar in view of Lee teaches the target node of claim 19, wherein obtaining a second list of features used by the second source domain comprises:
sending to a second source domain a second feature list request; and receiving, in response to the second feature list request, a second list of features used by the second source domain (Jawahar, Par. [0041], “In an embodiment, the data processing server 104 may be configured to receive the classification request from the user-computing device 102 for classification of the plurality of unlabeled instances of the target domain. In an embodiment, the classification request may include the plurality of unlabeled instances of the target domain and the plurality of labeled instances of the source domain.”, therefore, a request may be sent to a second source domain (user inquiring classification regarding a second source domain) and in response, a second list of features/labeled instances of the source domain may be received).
Regarding Claim 22, Jawahar in view of Lee teaches the target node of claim 18, further comprising: obtaining the first decoder model (Lee, Claim 1, “generating a first neural network translation model which includes a neural network having an encoder-decoder structure and learns a feature of source domain data used in an unspecific field;”, thus, a first decoder model is obtained).
The reasons of obviousness have been noted in the rejection of Claim 18 above and applicable herein.
Regarding Claim 23, Jawahar in view of Lee teaches the target node of claim 22, wherein obtaining the first decoder model comprises:
requesting the first decoder model from the first source domain (Jawahar, Par. [0040], “In an embodiment, the user may utilize the user-computing device 102 to transmit a classification request to the data processing server 104 for classification of a plurality of unlabeled instances of a target domain.”, thus, a request may be transmitted regarding a classification for the first decoder model from the first source domain. Although there is no explicit recitation of the first source domain in this citation, as mentioned in the rejection of the Independent claims above, Jawahar teaches transfer learning for a plurality of source domains – hence, this plurality comprises a first/second source domain); and receiving the first decoder model (Lee, Claim 1, “generating a first neural network translation model which includes a neural network having an encoder-decoder structure and learns a feature of source domain data used in an unspecific field;”, thus, a first decoder model is received).
The reasons of obviousness have been noted in the rejection of Claim 18 above and applicable herein.
Regarding Claim 24, Jawahar in view of Lee teaches the target node of claim 18, further comprising:
obtaining the second decoder (Lee, Claim 1, “generating a second neural network translation model which includes a neural network having the encoder-decoder structure and learns a feature of target domain data used in a specific field;”, thus, a second decoder model is obtained) model, wherein the second decoder model has been trained by the second source domain conditionally on the second set of target features (Jawahar, Par. [0099], “In an embodiment, the adaptation processor 208 may be configured to formulate a second neural network based on the pseudo-labeled instances of the target domain and the learned common representation. The pseudo-labeled instances of the target domain may constitute a second input layer of the second neural network and the learned common representation may constitute a second hidden layer of the second neural network.”, therefore, the second neural network model has been trained by the second source domain conditionally on the subset of common features/common representation).
The reasons of obviousness have been noted in the rejection of Claim 18 above and applicable herein.
Regarding Claim 26, Jawahar in view of Lee teaches the target node of claim 24, wherein obtaining the second decoder model comprises:
requesting the second decoder model from the second source domain (Jawahar, Par. [0040], “In an embodiment, the user may utilize the user-computing device 102 to transmit a classification request to the data processing server 104 for classification of a plurality of unlabeled instances of a target domain.”, thus, a request may be transmitted regarding a classification for the second decoder model from the second source domain. Although there is no explicit recitation of the second source domain in this citation, as mentioned in the rejection of the Independent claims above, Jawahar teaches transfer learning for a plurality of source domains – hence, this plurality comprises a first/second source domain); and receiving the second decoder model (Lee, Claim 1, “generating a second neural network translation model which includes a neural network having the encoder-decoder structure and learns a feature of target domain data used in a specific field;”, thus, a second decoder model is received).
The reasons of obviousness have been noted in the rejection of Claim 18 above and applicable herein.
Regarding Claim 27, Jawahar in view of Lee teaches the target node of claim 18, further comprising determining a decoder order sequence based on a number of features that are common among two or more source domains, wherein the decoder order sequence indicates an order in which to generate the first data and the second data (Lee, Par. [0077], “The decoder 256 may sequentially decode the common feature vector values in the order, in which words constituting the common feature are sorted, to output a common feature-based output vector value corresponding to a translation result of the common feature. The decoder 256 may be configured with a recurrent neural network or a convolutional neural network.”, therefore, a sequential order of decoders may be determined based on the common features).
The reasons of obviousness have been noted in the rejection of Claim 18 above and applicable herein.
Regarding Claim 28, Jawahar in view of Lee teaches the target node of claim 19, further comprising:
determining a number of features that are common among two or more source domains based on the first list of features and the second list of features ((Jawahar, Par. [0043], “The data processing server 104 may be further configured to learn a source specific representation corresponding to the source domain and a common representation shared between the source domain and the target domain based on the plurality of labeled instances of the source domain. […] The common representation may include a plurality of common features extracted from the plurality of labeled instances (such as a text segment, an audio segment, and/or the like) of the source domain that are used in the source domain as well as in the target domain.”, therefore, a common representation, representing common features between both models and their according source domains, is determined based on the first/second features) and
determining a decoder order sequence based on the number of features that are common among the two or more source domains, wherein the decoder order sequence indicates an order in which to generate the first data and the second data (Lee, Par. [0077], “The decoder 256 may sequentially decode the common feature vector values in the order, in which words constituting the common feature are sorted, to output a common feature-based output vector value corresponding to a translation result of the common feature. The decoder 256 may be configured with a recurrent neural network or a convolutional neural network.”, therefore, a sequential order of decoders may be determined based on the common features).
The reasons of obviousness have been noted in the rejection of Claim 18 above and applicable herein.
Regarding Claim 30, Jawahar in view of Lee teaches the target node of claim 18, wherein generating a first data by using the first decoder (See introduction of Lee reference above which teaches a first decoder model) model with the first set of target features comprises filtering data generated by the first decoder model based on a similarity between source and target features (Jawahar, Par. [0043], “The data processing server 104 may be further configured to learn a source specific representation corresponding to the source domain and a common representation shared between the source domain and the target domain based on the plurality of labeled instances of the source domain. […] The common representation may include a plurality of common features extracted from the plurality of labeled instances (such as a text segment, an audio segment, and/or the like) of the source domain that are used in the source domain as well as in the target domain.”, therefore, the common representation/similarity between source and target features is used in the generation of first data/classification output in source domain – this is better depicted by Figure 4 which depicts the first neural network (label 402) which considers the common representation (label 406B) to generate first data/classification output (label 408)); and
wherein generating a second data by using the second decoder (See introduction of Lee reference above which teaches a second decoder model) model with the second set of target features comprises filtering data generated by the second decoder model based on the similarity between source and target features (Jawahar, Par. [0043], “The data processing server 104 may be further configured to learn a source specific representation corresponding to the source domain and a common representation shared between the source domain and the target domain based on the plurality of labeled instances of the source domain. […] The common representation may include a plurality of common features extracted from the plurality of labeled instances (such as a text segment, an audio segment, and/or the like) of the source domain that are used in the source domain as well as in the target domain.”, therefore, the common representation/similarity between source and target features is used in the generation of first data/classification output in source domain – this is better depicted by Figure 4 which depicts the second neural network (label 412) which considers the common representation (label 406B) to generate first data/classification output (label 418)).
The reasons of obviousness have been noted in the rejection of Claim 18 above and applicable herein.
Regarding Claim 31, Jawahar in view of Lee teaches the target node of claim 30, wherein similarity between source and target features is determined based on one or more distance measures (Jawahar, Par. [0045], “The merging may be based on a similarity score between the first target specific feature and the second target specific feature. The data processing server 104 may use one or more similarity measures known in the art for the generation of the similarity score.”, therefore, the similarity between features is determined based on one or more distance measures, including a similarity score).
Regarding Claim 32, Jawahar in view of Lee teaches the target node of claim 31, wherein the one or more distance measures are selected from the group consisting of a cosine similarity measure, a K-L divergence measure, a Euclidean measure, a Wasserstein measure, and a dot-product measure (Jawahar, Par. [0045], “Examples of the one or more similarity measures known in the art for the generation of similarity score may include, but are not limited to, cosine similarity measure and Lavenshtein similarity measure.”, therefore, the distance measure may comprise a cosine similarity measure).
Regarding Claim 33, Jawahar in view of Lee teaches the target node of claim 18, further comprising:
sending to a third source domain a third feature list request (Jawahar, Par. [0041], “In an embodiment, the data processing server 104 may be configured to receive the classification request from the user-computing device 102 for classification of the plurality of unlabeled instances of the target domain. In an embodiment, the classification request may include the plurality of unlabeled instances of the target domain and the plurality of labeled instances of the source domain. In another embodiment, the classification request may not include the plurality of labeled instances of the source domain.”, thus, a request may be sent for another set of features, pertaining to another source domain. Further, Par. [0115] mentions that the transfer learning is based on a plurality of source domains, thus, the request may be sent to a third source domain);
receiving, in response to the third feature list request, a third list of features used by the third source domain (Jawahar, Par. [0041], “In another embodiment, the classification request may not include the plurality of labeled instances of the source domain. In such a scenario, the data processing server 104 may be configured to retrieve the labeled instances of the source domain from one or more social media websites or the database server 106. The unlabeled instances of the target domain and the labeled instances of the source domain may correspond to real-time input data, such that the unlabeled instances of the target domain and the labeled instances of the source domain may be received in real time or near real time from the user-computing device 102.”, therefore, a third list of features/instances pertaining to the relevant third source domain may be received in response to such a request);
requesting (Jawahar, Par. [0040], “In an embodiment, the user may utilize the user-computing device 102 to transmit a classification request to the data processing server 104 for classification of a plurality of unlabeled instances of a target domain.”, therefore, a user may request another model to perform classification) a third decoder model with a third set of target features from the third source domain, wherein the third set of target features comprises the third list of features (Lee, Claim 1, “generating a third neural network translation model which includes a neural network having the encoder-decoder structure and learns a common feature of the source domain data and the target domain data;”, therefore, a third neural network comprising a third decoder is disclosed. Further, the third decoder model may comprise a third set of target features/third list of features which are common between the source and target domain);
receiving the third decoder model, wherein the third decoder model has been trained by the third source domain conditionally on the subset of features common to the third set of target features and both the first and second sets of target features (Lee, Claim 1, “generating a third neural network translation model which includes a neural network having the encoder-decoder structure and learns a common feature of the source domain data and the target domain data; and”, thus, a third decoder model is received and has been trained conditionally on the subset of features common between the source domain data and target domain data);
generating a third data by using the third decoder model with the third set of target features (Lee, Claim 4, “The neural network translation model constructing method of claim 1, wherein the generating of the third neural network translation model comprises: generating an encoder outputting a common feature vector value obtained by encoding the common feature; generating a domain classifier classifying which of the source domain and the target domain the common feature vector value is included in; and generating a decoder which decodes the common feature vector value classified by the domain classifier to output an output vector value corresponding to a translation result of the common feature.”, thus, third data/output data may be generated by using the third decoder model with the third set of target features (common features)); and
updating the final set of target features and final data based on the generated third data (Lee, Claim 2, “The neural network translation model constructing method of claim 1, wherein based on a combination operation of the combiner, the neural network translation model functions as an ensemble model obtained by a combination of the translation results of the first to third neural network translation models.”, thus, the final set of target features and data may be updated based on the generated third data, as the data is combined between decoders to provide a final result).
The reasons of obviousness have been noted in the rejection of Claim 18 above and applicable herein.
10. Claim 29 is rejected under 35 U.S.C. 103 as being unpatentable over Jawahar et al. (hereinafter Jawahar) (US PG-PUB 20180218284), in view of Lee et al. (hereinafter Lee) (US PG-PUB 20190114545), further in view of Mansi et al. (hereinafter Mansi) (US PG-PUB 20200034654).
Regarding Claim 29, Jawahar in view of Lee teaches the target node of claim 18.
Jawahar in view of Lee does not explicitly disclose wherein one or more of the first decoder model and the second decoder model are one of a conditional Generative Adversarial Network (GAN) type model and a conditional Variational Autoencoder (VAE) type model.
However, Mansi teaches wherein one or more of the first decoder model and the second decoder model are one of a conditional Generative Adversarial Network (GAN) type model and a conditional Variational Autoencoder (VAE) type model (Mansi, Par. [0052], “In some examples of multimodal registration, the style transfer probability density is modelled using a conditional generative model. Examples of suitable conditional generative models are those generated by conditional variational autoencoders (CVAEs) and conditional generative adversarial networks (CGANs). FIG. 6 shows an example of a conditional variational autoencoder (CVAE) 600 for generating a conditional model in accordance with the present embodiment. The CVAE 600 includes an encoder network 602 and a decoder network 604.”, thus, the first/second decoder models (See Mansi Par. [0040-0041] which explicitly recites the use of decoders for style transfer learning between networks) may be one of a conditional generative adversarial network (GAN) type model and a conditional variational autoencoder (VAE) type model).
It would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the target node of claim 18, as disclosed by Jawahar in view of Lee to include wherein one or more of the first decoder model and the second decoder model are one of a conditional Generative Adversarial Network (GAN) type model and a conditional Variational Autoencoder (VAE) type model, as disclosed by Mansi. One of ordinary skill in the art would have been motivated to make this modification to enable the use of conditional models, which may improve performance and accuracy of the system even when handling unseen data (Mansi, Par. [0072], “Updating parameters of the generative model and/or the conditional model using image data as described above results in the prior probability distribution and/or the conditional probability distribution more accurately reflecting the data as the method is performed. Using large volumes of image data, and performing the refinement steps iteratively, the models may become highly accurate even in cases where the generative model and/or the conditional model were poorly pretrained, for example due to a lack of ground-truth data or sufficiently accurate synthetic data.”).
Conclusion
11. 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.
12. Any inquiry concerning this communication or earlier communications from the examiner should be directed to Devika S Maharaj whose telephone number is (571)272-0829. The examiner can normally be reached Monday - Thursday 8:30am - 5:30pm.
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, Alexey Shmatov can be reached at (571)270-3428. 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.
/DEVIKA S MAHARAJ/Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights