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 .
Response to Remarks
Claim Rejections – 35 U.S.C. 101
Applicant’s amendments have been fully considered but they are not persuasive.
Applicant argues (pg. 11) that “all of the pending claims are patent eligible under at least Step 2A, prong one” because “inference model training is not an abstract idea” according to the August 4 Memo published by the USPTO.
Examiner respectfully disagrees. While inference model training in it of itself isn’t necessarily an abstract idea, the other elements in the limitations are. Taken as a whole, they are determined to be a mental process. See rejection below for further details.
Applicant notes (pg. 11) that in the 101-rejection portion of the nonfinal rejection Office Action, Examiner omitted the Step 2A, prong two of the Office’s Subject Matter Eligibility Test.
Examiner respectfully disagrees. Examiner did include the Step 2A, prong two analysis, either explicitly or implicitly for every claim. For the claims where this is not explicitly stated, it was implied that the claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception. This is now explicitly mentioned in the claims for clarification. See rejection below for details.
Applicant argues (pg. 11-12) that the Specification discloses that the invention reduces the quantity of data transmitted from the data collector to the data aggregator, and as a result, frees up computing resources of the data collectors to be redirected to be used for other essential processes.
Examiner respectfully disagrees. While the Specification describes an improvement in the efficiency of the computing system of the data collector, the claims do not fully reflect this improvement, in particular with respect to the training/retraining of the twin inference model. The amended limitations do not reflect this improvement in regard to the machine learning model, as the training/retraining appears to be a black box based on just the claims. Examiner suggests expanding on this improvement with relation to the machine learning model.
The foregoing applies to all independent claims and their dependent claims.
Claim Rejections – 35 U.S.C. 102, 103
Applicant’s prior art arguments have been fully considered and they are persuasive.
Applicant argues (pg. 13) that the cited references do not teach the newly amended limitations that further clarify retraining the inference models.
Examiner agrees. Accordingly, a new reference, which was previously referenced in the Conclusion of the nonfinal rejection, Velagapudi (US 20210232911 A1) has been added to the rejection, as further detailed below.
The foregoing applies to all independent claims and their dependent claims.
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.
Claim 21 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention.
Claim 21 recites the limitation "non-date reduced state" in line 2. There is insufficient antecedent basis for this limitation in the claim. The phrase “non-date reduced state”, is not defined by the claim, nor is mentioned in the specification. For examination purposes the examiner will take this as a typo and treat it as not a sized-reduced state.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Claims 1-10 are method claims. Claims 11-20 are machine/system/product claims. Therefore, claims 1-20 are directed to either a process, machine, manufacture or composition of matter.
With respect to claim 1:
Step 2A – Prong 1:
…
…
…
…
using, … the representation of the data as a substitute for a normal size version of the first data collected by the data collector over the predetermined period of time without ever receiving a full copy of the normal size version of the first data from the data collector; (mental process – a person can recognize the use of the representation of the data as a substitute for a normal size version of the first data collected by the data collector over the predetermined period of time without ever receiving a full copy of the normal size version of the first data from the data collector.)
revising, … and after obtaining the representation of the data, the representation of the data using subsequently collected data from the data collector to obtain a revised representation of the data, the subsequently collected data being obtained via a transmission from the data collector, … (mental process – a person can manually revise the representation of the data using subsequently collected data from the data collector, with the assistance of a pen/paper.)
and performing an action set using the revised representation of the data. (mental process – a person can manually perform an action set based on the revised reconstructed data, with the assistance of a pen/paper.)
Step 2A – Prong 2: This judicial exception is not integrated into a practical application.
A method for managing data collection in a distributed system where data is collected in a data aggregator of the distributed system and from a data collector of the distributed system that is operably connected to the data aggregator via a communication system, the method comprising: (mere instructions to apply the exception using a generic computer component – data aggregator and communication system apply exception.)
obtaining, by the data aggregator, reduced size data from the data collector, the reduced sized data being a reduced size version of first data collected by the data collector over a predetermined period of time; (Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g)).
generating, by the data aggregator using a local copy of a twin inference model hosted by the data aggregator, a locally generated inference duplicative of an inference upon which the reduced size data is based; (Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g)).
generating, by the data aggregator, a representation of data upon which the reduced size data is based using: the reduced size data, and the locally generated inference; (Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g)).
… by the data aggregator, … (mere instructions to apply the exception using a generic computer component – data aggregator applies exception.)
… by the data aggregator, … the revising comprising at least a retraining of the local copy of the twin inference model using the subsequently collected data from the data collector; (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 retraining of the local copy of the twin inference model.);
…
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
A method for managing data collection in a distributed system where data is collected in a data aggregator of the distributed system and from a data collector of the distributed system that is operably connected to the data aggregator via a communication system, the method comprising: (mere instructions to apply the exception using a generic computer component – data aggregator and communication system apply exception.)
obtaining, by the data aggregator, reduced size data from the data collector, the reduced sized data being a reduced size version of first data collected by the data collector over a predetermined period of time; (MPEP 2106.05(d)(II) indicate that merely “Receiving or transmitting data over a network, e.g., using the Internet to gather data” is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim – the reduced size data is merely received from the data collector). Thereby, a conclusion that the claimed distribute step is well-understood, routine, conventional activity is supported under Berkheimer.)
generating, by the data aggregator using a local copy of a twin inference model hosted by the data aggregator, a locally generated inference duplicative of an inference upon which the reduced size data is based; (MPEP 2106.05(d)(II) indicate that merely “Receiving or transmitting data over a network, e.g., using the Internet to gather data” is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim – the a locally generated inference duplicative of an inference upon which the reduced size data is based is merely received). Thereby, a conclusion that the claimed distribute step is well-understood, routine, conventional activity is supported under Berkheimer.)
generating, by the data aggregator, a representation of data upon which the reduced size data is based using: the reduced size data, and the locally generated inference; (MPEP 2106.05(d)(II) indicate that merely “Receiving or transmitting data over a network, e.g., using the Internet to gather data” is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim – the representation of data upon which the reduced size data is based is merely received). Thereby, a conclusion that the claimed distribute step is well-understood, routine, conventional activity is supported under Berkheimer.)
… by the data aggregator, … (mere instructions to apply the exception using a generic computer component – data aggregator applies exception.)
… by the data aggregator, … the revising comprising at least a retraining of the local copy of the twin inference model using the subsequently collected data from the data collector; (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 retraining of the local copy of the twin inference model.);
…
With respect to claim 2:
Step 2A – Prong 1:
…
…
…
…
reconstructing a second representation of the data upon which the reduced size data is based using the reduced size data and the revised locally generated inference; (mental process – a person can manually reconstruct a second representation of the data upon which the reduced size data is based using the reduced size data and the revised locally generated inference with the assistance of a pen/paper.)
and updating the representation of the data using the second representation of the data to obtain the revised representation of the data. (mental process – a person can manually update the representation of the data using the second representation of the data with the assistance of a pen/paper.)
Step 2A – Prong 2: This judicial exception is not integrated into a practical application.
The method of claim 1, wherein revising the representation of the data comprises: obtaining a data sample of the subsequently collected data; (Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g)).
obtaining an updated inference model using the data sample … (Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g)).
… by retraining the local copy of the twin inference model hosted by the data aggregator using the data sample; (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 retraining of the local copy of the twin inference model.);
obtaining a revised locally generated inference using the updated inference model; (Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g)).
…
…
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The method of claim 1, wherein revising the representation of the data comprises: obtaining a data sample of the subsequently collected data; (MPEP 2106.05(d)(II) indicate that merely “Receiving or transmitting data over a network, e.g., using the Internet to gather data” is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim – the data sample of the subsequently collected data is merely received). Thereby, a conclusion that the claimed distribute step is well-understood, routine, conventional activity is supported under Berkheimer.)
obtaining an updated inference model using the data sample … (MPEP 2106.05(d)(II) indicate that merely “Receiving or transmitting data over a network, e.g., using the Internet to gather data” is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim – the updated inference model based on the local copy of the twin inference model is merely received). Thereby, a conclusion that the claimed distribute step is well-understood, routine, conventional activity is supported under Berkheimer.)
… by retraining the local copy of the twin inference model hosted by the data aggregator using the data sample; (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 retraining of the local copy of the twin inference model.);
obtaining a revised locally generated inference using the updated inference model; (MPEP 2106.05(d)(II) indicate that merely “Receiving or transmitting data over a network, e.g., using the Internet to gather data” is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim – the revised locally generated inference is merely received). Thereby, a conclusion that the claimed distribute step is well-understood, routine, conventional activity is supported under Berkheimer.)
…
…
With respect to claim 3:
Step 2A – Prong 1:
The method of claim 2, wherein the local copy of the twin inference model comprises a neural network. (mental process – a person can recognize that the local copy of the twin inference model comprises a neural network.)
Step 2A – Prong 2: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 4:
Step 2A – Prong 1:
The method of claim 3, wherein the updated inference model is obtained by retraining the local copy of the twin inference model using the data sample. (mental process – a person can recognize that the updated inference model is obtained by retraining the local copy of the twin inference model using the data sample.)
Step 2A – Prong 2: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 5:
Step 2A – Prong 1:
The method of claim 2, wherein the representation of the data is different from the normal size version of the first data collected by the data collector over the predetermined period by a first difference amount consistent with a level of inaccuracy of the locally generated inference. (mental process – a person can recognize that the representation of the data is different from the normal size version of the first data collected by the data collector over the predetermined period by a first difference amount consistent with a level of inaccuracy of the locally generated inference.)
Step 2A – Prong 2: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 6:
Step 2A – Prong 1:
The method of claim 5, wherein the revised representation of the data is more similar to the normal size version of the first data collected by the data collector over the predetermined period than the representation of the data as a result of a second level of inaccuracy of the revised locally generated inference being smaller than the level of inaccuracy of the locally generated inference. (mental process – a person can recognize that the revised representation of the data is more similar to the normal size version of the first data collected by the data collector over the predetermined period than the representation of the data as a result of a second level of inaccuracy of the revised locally generated inference being smaller than the level of inaccuracy of the locally generated inference.)
Step 2A – Prong 2: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 7:
Step 2A – Prong 1:
The method of claim 1, wherein the reduced size data indicates that the locally generated inference is a sufficiently accurate representation of the first data collected by the data collector over the predetermined period of time such the fully copy of the normal size version of the first data collected by the data collector over the predetermined period of time is not required to be provided to the data aggregator by the data collector. (mental process – a person can recognize that the reduced size data indicates that the locally generated inference is a sufficiently accurate representation of the first data collected by the data collector over the predetermined period of time such the fully copy of the normal size version of the first data collected by the data collector over the predetermined period of time is not required to be provided to the data aggregator by the data collector.)
Step 2A – Prong 2: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 9:
Step 2A – Prong 1:
The method of claim 7, wherein the representation of the data upon which the reduced size data is based on is the locally generated inference. (mental process – a person can recognize that the representation of the data upon which the reduced size data is based is the locally generated inference.)
Step 2A – Prong 2: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
With respect to claim 10:
Step 2A – Prong 1:
…
and replacing, by the data aggregator and after the revising of the representation of the data, the representation of the data stored in the storage with the revised representation of the data. (mental process – a person can manually replace the stored representation of the data upon which the reduced size data is based with the revised representation of the data with the assistance of a pen/paper.)
Step 2A – Prong 2: This judicial exception is not integrated into a practical application.
The method of claim 9, further comprising: storing, by the data aggregator and in a storage, the representation of the data upon which the reduced size data is based as the substitute for the copy of first data collected by the data collector over the predetermined period of time; (Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g)).
Step 2B: The claim does not include additional elements considered individually and in combination that are sufficient to amount to significantly more than the judicial exception.
The method of claim 9, further comprising: storing, by the data aggregator and in a storage, the representation of the data upon which the reduced size data is based as the substitute for the copy of first data collected by the data collector over the predetermined period of time; (MPEP 2106.05(d)(II) indicate that merely “Storing and retrieving information in memory” is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is in the present claim – the representation of the data upon which the reduced size data is based is merely stored). Thereby, a conclusion that the claimed distribute step is well-understood, routine, conventional activity is supported under Berkheimer.)
Claim 11 is substantially similar to claim 1, but has the following additional elements:
With respect to claim 11:
Step 2A – Prong 2: This judicial exception is not integrated into a practical application.
A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations for managing data collection in a distributed system where data is collected in a data aggregator of the distributed system and from a data collector of the distributed system that is operably connected to the data aggregator via a communication system, the operations comprising: (mere instructions to apply the exception using a generic computer component – processor, data aggregator, and communication system apply exception.)
Claims 12-15 are rejected on the same grounds under 35 U.S.C. 101 as claims 2-5, as they are substantially similar, respectively. Mutatis mutandis.
Claim 16 is substantially similar to claim 1, but has the following additional elements:
With respect to claim 16:
Step 2A – Prong 2: This judicial exception is not integrated into a practical application.
A data aggregator, comprising: a processor; (mere instructions to apply the exception using a generic computer component – processor applies exception.)
and a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations for managing data collection in a distributed system where data is collected in a data aggregator of the distributed system and from a data collector of the distributed system that is operably connected to the data aggregator via a communication system, the operations comprising: (mere instructions to apply the exception using a generic computer component – memory applies exception.)
Claims 17-20 are rejected on the same grounds under 35 U.S.C. 101 as claims 2-5, as they are substantially similar, respectively. Mutatis mutandis.
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, 11-14, 16-19, 21 are rejected under 35 U.S.C. 103 as being unpatentable over Stergioudis (US 20220201008 A1) hereinafter known as Stergioudis in view of Velagapudi et al. (US 20210232911 A1) hereinafter known as Velagapudi.
Regarding independent claim 1, Stergioudis teaches:
A method for managing data collection in a distributed system where data is collected in a data aggregator of the distributed system and from a data collector of the distributed system that is operably connected to the data aggregator via a communication system, the method comprising: obtaining, by the data aggregator, reduced size data from the data collector, the reduced sized data being a reduced size version of first data collected by the data collector over a predetermined period of time; (Stergioudis [¶ 0061]: “The client device 201 can receive inputs from the user in an attempt to signed-in.” Stergioudis teaches a client device that collects information regarding inputs from the user to attempt a sign-in. Stergioudis [¶ 0062]: “The network 203 can facilitate a session of an application running on the client device 201 to transmit to or receive data from the DPS 202” Stergioudis teaches a data processing system (DPS) that receives data from the client device through a network. The DPS aggregates data from the client device. Stergioudis [¶ 0063]: “For example, the network 203 can compress, reformat, convert, or otherwise forward the data packages from a device to another.” Stergioudis teaches that the data from the client device can be compressed, or reduced, prior to it being received by the DPS. Stergioudis [¶ 0007]: “Features of the collected information can be extracted to identify any patterns of the user behaviors throughout a session (e.g., an application session)” Stergioudis teaches that the collected information is from a session, which is a predetermined length of an application session.)
generating, by the data aggregator using a local copy of a twin inference model hosted by the data aggregator, a locally generated inference duplicative of an inference upon which the reduced size data is based; (Stergioudis [Figure 5B]: Stergioudis teaches that the samples of data are fed into two neural networks. These two networks have shared weights, making them a pair of networks, or twin networks. These pair of networks each output a representation of the user. Stergioudis [¶ 0043]: “This representation can refer to the behavior of the user which can deviate from the common path that the account owner follows.” Stergioudis teaches that the representation is a behavior of the user that has some variance from the typical behavior of the user. This shows that the models are inference models, as the representations infer the behavior of the user, with some margin of error. Stergioudis [¶ 0179]: “The system can access a data repository storing multiple distance models” Stergioudis teaches that the data repository stores multiple, which in the case of the twin models is two, models that determine the distance from the behavior of the client to their historical behavior.)
generating, by the data aggregator, a representation of data upon which the reduced size data is based using: the reduced size data, and the locally generated inference; (Stergioudis [¶ 0043]: “This representation can refer to the behavior of the user which can deviate from the common path that the account owner follows.” Stergioudis teaches that the representation is a behavior of the user that has some variance from the typical behavior of the user. This inference is based on the reduced size data of the user input from the client device.)
using, by the data aggregator, the representation of the data as a substitute for a normal size version of the first data collected by the data collector over the predetermined period of time without ever receiving a full copy of the normal size version of the first data from the data collector; (Stergioudis [¶ 0063]: “For example, the network 203 can compress, reformat, convert, or otherwise forward the data packages from a device to another.” Stergioudis teaches that the data from the client device can be compressed, or reduced, prior to it being received by the DPS. Provided that this is the standard for data sharing between the two entities, or if it’s the only instance of data sharing, then the receiver will never receive an uncompressed version of the data.)
revising, by the data aggregator and after obtaining the representation of the data, the representation of the data using subsequently collected data from the data collector to obtain a revised representation of the data, the subsequently collected data being obtained via a transmission from the data collector, … (Stergioudis [¶ 0066]: “The analytics service 208 can process sequences of categorical events to determine user behaviors. The sequences of categorical events can indicate the historical patterns of the user. For example, the analytics service 208 can receive location data and network connection data of the client device 201 as an input feature. The device owner may not commonly use public Wi-Fi, for example. The analytics service 208 can process sequences of event types produced by the user, such as file download, file upload, folder delete, or folder created within a session.” Stergioudis teaches that the analytics service receives sequential data from the client device. This shows that a collection of the sequential data is updated by a subsequent update of the data. This revises the representation of the user, as further information, such as the event types that the user commonly produces, updates their profile.)
and performing an action set using the revised representation of the data. (Stergioudis [¶ 0066]: “The analytics service 208 can determine the user behavior is unusual based on a comparison with the user historical patterns. For instance, the analytics service 208 can use a historical sequence of locations to detect whether the current location of the user (e.g., corresponding to the location of the client device 201) is expected or not.” Stergioudis teaches that in response to the subsequent update of the data, the representation of the user is revised and updated to determine if a behavior is in line with their historical patterns. For instance, an assessment can be made on whether the current location of the user is expected or not, based on this update of data.)
Stergioudis does not explicitly teach:
… the revising comprising at least a retraining of the local copy of the twin inference model using the subsequently collected data from the data collector;
However, Velagapudi teaches:
… the revising comprising at least a retraining of the local copy of the twin inference model using the subsequently collected data from the data collector; (Velagapudi [¶ 0060]: “The user may wish to retrain the embedding and/or neural networks when switching from analyzing training data associated from a first domain to analyzing training data associated with a second domain.” Velagapudi teaches that of the different domains of how the machine learning model is to be used (in terms of context), the user may retrain the model with data from the second domain after it has trained using the data from the first domain.)
Stergioudis and Velagapudi are in the same field of endeavor as the present invention, as the references are directed to training twin models involving reducing data and retraining using data from different context, respectively. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine collecting data and making inferences on it using a twin model as taught in Stergioudis with retraining the model using data from a separate context as taught in Velagapudi. Velagapudi provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Stergioudis to include teachings of Velagapudi because the combination would allow for the model to be applicable to various different contexts of data, increasing its ability to make accurate inferences.
Regarding dependent claim 2, Stergioudis and Velagapudi teach:
The method of claim 1, wherein revising the representation of the data comprises: obtaining a data sample of the subsequently collected data; (Stergioudis [¶ 0066]: “The analytics service 208 can process sequences of categorical events to determine user behaviors. The sequences of categorical events can indicate the historical patterns of the user. For example, the analytics service 208 can receive location data and network connection data of the client device 201 as an input feature.” Stergioudis teaches that the samples of data are collected in subsequent timings – for instance, the location data of the user data is a sequence of data for different points in time.)
obtaining an updated inference model using the data sample … (Stergioudis [¶ 0013]: “The method can include one or more processors training the distance model with the historical data for one or more entities using a distance-based loss function configured to predict relative distances between inputs for the one or more entities.” Stergioudis teaches that using the historical data, the model is trained to predict the distance between inputs for a user. Stergioudis [¶ 0182]: The system can input the data into the distance model to generate the representation. The system can identify a first historical representation based on a first data sample of the historical data input into the distance model. Stergioudis teaches that the distance model, which as shown above may be a twin model, is used to obtain the updated prediction/inference that is the distance between input samples for a user.)
… by retraining the local copy of the twin inference model hosted by the data aggregator using the data sample; (Velagapudi [¶ 0060]: “The user may wish to retrain the embedding and/or neural networks when switching from analyzing training data associated from a first domain to analyzing training data associated with a second domain.” Velagapudi teaches that of the different domains of how the machine learning model is to be used (in terms of context), the user may retrain the model with data from the second domain after it has trained using the data from the first domain.)
obtaining a revised locally generated inference using the updated inference model; (Stergioudis [¶ 0179]: “The system can access a data repository storing multiple distance models” Stergioudis teaches that the data repository stores models that determine the distance from the behavior of the client to their historical behavior. This locally stored model gives the locally generated inference regarding the distance.)
reconstructing a second representation of the data upon which the reduced size data is based using the reduced size data and the revised locally generated inference; (Stergioudis [¶ 0182]: The system can identify a second historical representation based on a second data sample of the historical data input into the distance model. Stergioudis teaches that the second data sample, which as shown above may be compressed in size compared to the first data sample, is used to identify a second historical representation using the distance model.)
and updating the representation of the data using the second representation of the data to obtain the revised representation of the data. (Stergioudis [¶ 0109]: “The analytics service 208 can update an existing historical representation of the user based on the historical data or the features extracted from the historical data. The current representation of the user can be a historical representation for the next iteration to determine a second distance of the user.” Stergioudis teaches that the process for training the distance model and the process of updating the representation of the data is iterative, and therefore updates with each iteration.)
The reasons to combine are substantially similar to those of claim 1.
Regarding dependent claim 3, Stergioudis teaches:
The method of claim 2, wherein the local copy of the twin inference model comprises a neural network. (Stergioudis [Figure 5B]: Stergioudis teaches that the samples of data are fed into two neural networks. These two networks have shared weights, making them a pair of networks, or twin networks. These pair of networks each output a representation of the user.)
The reasons to combine are substantially similar to those of claim 1.
Regarding dependent claim 4, Stergioudis teaches:
The method of claim 3, wherein the updated inference model is obtained by retraining the local copy of the twin inference model using the data sample. (Stergioudis [¶ 0109]: “The analytics service 208 can update an existing historical representation of the user based on the historical data or the features extracted from the historical data. The current representation of the user can be a historical representation for the next iteration to determine a second distance of the user.” Stergioudis teaches that the process for training the distance model and the process of updating the representation of the data is iterative, and therefore updates with each iteration.)
The reasons to combine are substantially similar to those of claim 1.
Independent claim 11 is rejected on the same grounds under 35 U.S.C. 103 as claim 1, as claim 11 is substantially similar to claim 1, but has the following additional elements:
Stergioudis teaches:
A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations for managing data collection in a distributed system where data is collected in a data aggregator of the distributed system and from a data collector of the distributed system that is operably connected to the data aggregator via a communication system, the operations comprising: (Stergioudis [¶ 0023]: “a non-transitory computer-readable medium comprising processor executable instructions that, when executed by at least one processor, cause the at least one processor to manage a session accessed by a client device.” Stergioudis teaches a non-transitory storage that stores instructions that are executable by a processor. Stergioudis [¶ 0023]: “The instructions can include instructions to receive data in a plurality of modalities corresponding to a plurality of features of the session for an entity accessed by the client device.” Stergioudis teaches that the instructions are for receiving data through a connection from a client device.)
The reasons to combine are substantially similar to those of claim 1.
Claims 12-14 are rejected on the same grounds under 35 U.S.C. 103 as claims 2-4 as they are substantially similar, respectively. Mutatis mutandis.
Independent claim 16 is rejected on the same grounds under 35 U.S.C. 103 as claim 1, as claim 16 is substantially similar to claim 1, but has the following additional elements:
Stergioudis teaches:
A data aggregator, comprising: a processor; (Stergioudis [¶ 0023]: “processor executable instructions that, when executed by at least one processor … receive data in a plurality of modalities corresponding to a plurality of features of the session for an entity accessed by the client device” Stergioudis teaches a processor that can execute instructions of aggregating data.)
and a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations for managing data collection in a distributed system where data is collected in a data aggregator of the distributed system and from a data collector of the distributed system that is operably connected to the data aggregator via a communication system, the operations comprising: (Stergioudis [¶ 0023]: “a non-transitory computer-readable medium comprising processor executable instructions that, when executed by at least one processor, cause the at least one processor to manage a session accessed by a client device.” Stergioudis teaches a non-transitory storage that stores instructions that are executable by a processor. Stergioudis [¶ 0023]: “The instructions can include instructions to receive data in a plurality of modalities corresponding to a plurality of features of the session for an entity accessed by the client device.” Stergioudis teaches that the instructions are for receiving data through a connection from a client device.)
The reasons to combine are substantially similar to those of claim 1.
Claims 17-19 are rejected on the same grounds under 35 U.S.C. 103 as claims 2-4 as they are substantially similar, respectively. Mutatis mutandis.
Regarding dependent claim 21, Stergioudis and Kim teach:
The method of claim 1, wherein the subsequently collected data obtained from the data collector in a non-date reduced state, and the data collector transmits the subsequently collected to the data aggregator in response to the data collector determining that the local copy of the twin inference model hosted by the data aggregator is unable to generate an accurate representation of the subsequently collected data. (Stergioudis [¶ 0009]: “The method can include one or more processors comparing the distance with a threshold established for the entity. The method can include one or more processors generating, based at least in part on the comparison between the distance with the threshold, an action to manage access by the client device to the session for the entity.” Stergioudis teaches that there is a threshold for the model, which is a measurement of the accuracy of the model and thus the collected data. If there is a lack of accuracy, information is sent to manage access of the client device.)
The reasons to combine are substantially similar to those of claim 1.
Claims 5-7, 9-10, 15, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Stergioudis in view of Velagapudi in view of Kim et al. (“Iterative learning-based many-objective history matching using deep neural network with stacked autoencoder”) hereinafter known as Kim.
Regarding dependent claim 5, Stergioudis and Velagapudi teach:
The method of claim 2,
Stergioudis and Velagapudi do not explicitly teach:
wherein the representation of the data is different from the normal size version of the first data collected by the data collector over the predetermined period by a first difference amount consistent with a level of inaccuracy of the locally generated inference.
However, Kim teaches:
wherein the representation of the data comprises a difference from the data upon which the reduced size data is based due to a level of inaccuracy of the locally generated inference. (Kim [Page 1469, Column 2, Paragraph 2]: “The dimension of the input data can be modified by the number of neurons in the hidden layer, the bottleneck, which is located in the middle of the SAE structure. If the number of neurons in the bottleneck is set smaller than the number of input data, the input data can be compressed (dimensionally reduced).” Kim teaches that by reducing the number of neurons in the hidden layer, the input data can be compressed. Kim [Page 1474, Figure 8]: Kim teaches that there is a level of inaccuracy of the model, measured as the mean reconstruction error, that depends on the number of hidden neurons.)
Kim is in the same field of endeavor as the present invention, since it is the analysis of inaccuracy of models based on the size of the data. It would have been obvious, before the effective filing date of the claimed invention, to a person of ordinary skill in the art, to combine collecting data and making inferences on it using a twin model as taught in Stergioudis as modified by Velagapudi with iteratively reducing the error of the model as the size of the data decreases as taught in Kim. Kim provides this additional functionality. As such, it would have been obvious to one of ordinary skill in the art to modify the teachings of Stergioudis as modified by Velagapudi to include teachings of Kim because the combination would allow for the twin model to iteratively train on increasingly reduced data. This has the potential benefit of decreasing the error of the model without overfitting to the data.
Regarding dependent claim 6, Stergioudis and Kim teach:
The method of claim 5,
Kim teaches:
wherein the revised representation of the data is more similar to the normal size version of the first data collected by the data collector over the predetermined period than the representation of the data as a result of a second level of inaccuracy of the revised locally generated inference being smaller than the level of inaccuracy of the locally generated inference. (Kim [Page 1474, Figure 8]: Kim teaches that there is a level of inaccuracy of the model, measured as the mean reconstruction error, that depends on the number of hidden neurons. The trend shows that the lower the number of hidden neurons, the lower the error rate is. In other words, the more compressed/reduced the data size is, the lower the level of inaccuracy becomes.)
The reasons to combine are substantially similar to those of claim 5.
Regarding dependent claim 7, Stergioudis teaches:
The method of claim 1,
Kim teaches:
wherein the reduced size data indicates that the locally generated inference is a sufficiently accurate representation of the first data collected by the data collector over the predetermined period of time such that the full copy of the normal size version of the first data collected by the data collector over the predetermined period of time is not required to be provided to the data aggregator by the data collector. (Kim [Page 1474, Figure 8]: Kim teaches that there is a level of inaccuracy of the model, measured as the mean reconstruction error, that depends on the number of hidden neurons. The trend shows that the lower the number of hidden neurons, the lower the error rate becomes. In the figure, the threshold where the error is sufficiently low such that the number of hidden neurons were not lowered further is at 200 neurons at just below 2% mean reconstruction error. As the decreasing of hidden neurons is analogous to the data transmitted to the data aggregator, as it is increasingly compressed data that is sent to be an input to the model, this is effectively a threshold to end the training.)
The reasons to combine are substantially similar to those of claim 5.
Regarding dependent claim 9, Stergioudis and Kim teach:
The method of claim 7,
Stergioudis teaches:
wherein the representation of the data upon which the reduced size data is based on is the locally generated inference. (Stergioudis [¶ 0043]: “This representation can refer to the behavior of the user which can deviate from the common path that the account owner follows.” Stergioudis teaches that the representation is a behavior of the user that has some variance from the typical behavior of the user. This inference is based on the reduced size data of the user input from the client device.)
The reasons to combine are substantially similar to those of claim 5.
Regarding dependent claim 10, Stergioudis and Kim teach:
The method of claim 9,
Stergioudis teaches:
further comprising: storing, by the data aggregator and in a storage, the representation of the data upon which the reduced size data is based as the substitute for the copy of first data collected by the data collector over the predetermined period of time; (Stergioudis [¶ 0074]: “Each representation can correspond to a feature extracted from a verified user data” Stergioudis teaches that the representations correspond to the features of the data. Stergioudis [¶ 0087]: “The machine learning engine 212 can output a representation for each data sample input into the model. Each representation can correspond to each data sample, such as a location representation, a file management representation, and a browsing representation of the user.” Stergioudis teaches that the representations correspond to the data samples, such as the location and browsing representations. This shows that the representations are effectively a copy of the data that it represents, in a different format.)
and replacing, by the data aggregator and after the revising of the representation of the data, the representation of the data stored in the storage with the revised representation of the data. (Stergioudis [¶ 0109]: “The analytics service 208 can update an existing historical representation of the user based on the historical data or the features extracted from the historical data.” Stergioudis teaches that the representation can be updated to be a new, revised representation that is based on the features from historical data.)
The reasons to combine are substantially similar to those of claim 5.
Claim 15 is rejected on the same grounds under 35 U.S.C. 103 as claim 5 as they are substantially similar. Mutatis mutandis.
Claim 20 is rejected on the same grounds under 35 U.S.C. 103 as claim 5 as they are substantially similar. Mutatis mutandis.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, 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 KYU HYUNG HAN whose telephone number is (703) 756-5529. The examiner can normally be reached on MF 9-5.
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 on (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.
/Kyu Hyung Han/
Examiner
Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123