CTNF 18/574,668 CTNF 92113 Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia 1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. DETAILED ACTION 2. This action is in response to the original filing on 12/27/2023. Claims 1-20 are pending and have been considered below. Information Disclosure Statement 3. The information disclosure statement (IDS(s)) submitted on 05/16/2024, 11/26/2025, 03/31/2026 is/are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner. Claim Rejections - 35 USC § 101 07-04-01 AIA 07-04 4. 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 the abstract idea without significantly more. Step 1 , the claims are directed to a process and manufacture. Step 2A Prong 1, Claims 1, 17, 18 recite, in part partitioning, based at least on the consent data for the subset of users, the set of users into a first group of users and a second group of users (Mental processes, a person could review consent information and place users into different categories) . Step 2A Prong 2 , this judicial exception is not integrated into a practical application. The additional elements: one or more computers; and one or more storage devices storing instructions that when executed by the one or more computers, cause the one or more computers to perform operations (mere instructions to apply the exception using a generic computer component). obtaining, for each user in a set of users, user data comprising user attribute data and, for a subset of the users, consent data for controlling usage of the user attribute data for the users in the subset of the users (mere data gathering and recited at a high level of generality, and thus are insignificant extra-solution activity). generating a first training dataset based on the user data for the first group of users (mere data gathering and recited at a high level of generality, and thus are insignificant extra-solution activity). generating a second training dataset based on the user data for the second group of users (mere data gathering and recited at a high level of generality, and thus are insignificant extra-solution activity). training, using the first training dataset and the second training dataset, a machine learning model configured to predict information about one or more users, the training comprising applying differential privacy to the second training dataset without applying differential privacy to the first training dataset (mere data gathering and output recited at a high level of generality, and thus are insignificant extra-solution activity). distributing, digital components to client devices using the machine learning model (field of use/ insignificant extra-solution activity). Step 2B , the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, either alone or in combination. The additional elements: one or more computers; and one or more storage devices storing instructions that when executed by the one or more computers, cause the one or more computers to perform operations (mere instructions to apply the exception using a generic computer component). obtaining, for each user in a set of users, user data comprising user attribute data and, for a subset of the users, consent data for controlling usage of the user attribute data for the users in the subset of the users (mere data gathering and recited at a high level of generality, and thus are insignificant extra-solution activity). generating a first training dataset based on the user data for the first group of users (mere data gathering and recited at a high level of generality, and thus are insignificant extra-solution activity). generating a second training dataset based on the user data for the second group of users (mere data gathering and recited at a high level of generality, and thus are insignificant extra-solution activity). training, using the first training dataset and the second training dataset, a machine learning model configured to predict information about one or more users, the training comprising applying differential privacy to the second training dataset without applying differential privacy to the first training dataset (mere data gathering and output recited at a high level of generality, and thus are insignificant extra-solution activity). distributing, digital components to client devices using the machine learning model (field of use/ insignificant extra-solution activity). Claims 2-16 and 19-20 provide further limitations to the abstract idea ( Mathematical concepts and/or Mental processes ) as rejected in claims 1, 17, 18, however, they do not disclose any additional elements that would amount to a practical application or significantly more than an abstract idea ( data gathering / insignificant extra-solution activity and/or generic computer component ). Claims 18 and 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. Claims 18 and 20 recite “computer-readable storage media”, however the specification is silent as to the particular details of what the “computer-readable storage media” comprises. It is noted that the ordinary meaning of the “computer-readable storage media” recited in the claim encompasses signals, carrier waves or other transmission media which is non-statutory. Thus, it is suggested to amend the “computer-readable storage media” recited in claims 18 and 20 to “non-transitory computer-readable storage media” to direct a/the computer-readable storage media to only physical media. Claim Rejections – 35 USC § 103 07-20-aia AIA 5. 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 of this title, 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 . 07-21-aia AIA 6. Claim s 1, 2, 5, 11, and 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Sanketi et al. (U.S. Patent Application Pub. No. US 20220004929 A1) in view of Amid et al. (Public Data-Assisted Mirror Descent for Private Model Training, arXiv, published 28 Mar 2022, pages 1-20) . Claim 1: Sanketi teaches a computer-implemented method comprising: obtaining, for each user in a set of users, user data comprising user attribute data and, for a subset of the users (i.e. the on-device machine learning platform can include a context provider that securely injects context features into collected training examples and/or client-provided input data used to generate predictions/inferences. Thus, the on-device machine learning platform can enable centralized training example collection, model training, and usage of machine-learned models as a service to applications or other clients … Example context features include: Audio State; Day Attributes; Calendar; Detected Activity; User-Specific Places (e.g., “home” vs. “work”; Network State; Power Connection; Screen Features; User Location; User Location Forecast; WiFi Scan Info; Weather; or other context features; para. [0028, 0036, 0135]) , consent data for controlling usage of the user attribute data for the users in the subset of the users (i.e. the training examples and context features described herein are simply provided for the purposes of illustrating example data that could be stored with training examples or used to provide inferences by the on-device platform. However, such data is not collected, used, or analyzed unless the user has provided consent after being informed of what data is collected and how such data is used. Further, the user can be provided with a tool to revoke or modify the scope of permissions. In addition, certain information or data can be treated in or more ways before it is stored or used, so that personally identifiable information is removed or stored in an encrypted fashion; para. [0039, 0040]) ; based at least on the consent data for the subset of users, the set of users into a first group of users and a second group of users (i.e. the on-device platform may obtain user permissions to all contexts listed above, but certain clients may not have the same permissions as the on-device platform has. So the platform can control the permissions for clients. In some implementations, this can be done by using a package manager to extract the permissions associated with a client application. The platform can maintain a mapping between contexts to permissions. Thus, a client will typically explicitly claim what contexts it wants to use when it registers itself with the on-device platform. The platform checks whether the client has the permissions to access the requested contexts. Only contexts to which the application has permission will be used to train and inference the corresponding model; para. [0040, 0151]) ; generating a training dataset based on the user data for the first group of users (i.e. the on-device machine learning platform 122 can securely inject context features descriptive of a context associated with the computing device 102 into the training examples. For example, upon receiving a training example from an application 120 a-c, a context provider component of the on-device platform 122 can determine one or more context features and can store such context features together with the training example in the centralized example database 124; para. [0032, 0068]) ; generating a training dataset based on the user data for the second group of users (i.e. the on-device machine learning platform 122 can securely inject context features descriptive of a context associated with the computing device 102 into the training examples. For example, upon receiving a training example from an application 120 a-c, a context provider component of the on-device platform 122 can determine one or more context features and can store such context features together with the training example in the centralized example database 124; para. [0032, 0068]) ; training, using the training dataset and the training dataset, a machine learning model configured to predict information about one or more users (i.e. After retraining of the model, the re-trained model can be used to provide inferences as described elsewhere herein. Typically, these inferences will have higher accuracy since the model has been re-trained on data that is specific to the user. Thus, the on-device machine learning platform can enable centralized example data collection and corresponding personalization of machine-learned models as a service to applications or other clients; para. [0038, 0044, 0076]) , the training comprising applying privacy to the training dataset (i.e. As another example technical effect and benefit, the on-device machine-learning platform can enable secure inclusion of contextual signals into training examples and/or inference inputs. That is, context features can be added to training examples or inference inputs in a manner than maintains privacy and complies with user-defined permissions. Through the inclusion of context information, the accuracy of the inferences provided by the machine-learned models can be improved … Only contexts to which the application has permission will be used to train and inference the corresponding model; para. [0051, 0151]) ; and distributing, digital components to client devices using the machine learning model (i.e. the applications 120 a-c can communicate with the on-device machine learning platform 122 via an API (which may be referred to as the “prediction API”) to provide input data and obtain predictions based on the input data from one or more of the machine-learned models 132 a-c. As an example, in some implementations, given a uniform resource identifier (URI) for a prediction plan (e.g., instructions for running the model to obtain inferences/predictions) and model parameters, the on-device machine learning platform 122 can download the URI content (e.g., prediction plan and parameters) and obtain one or more inferences/predictions by running the model (e.g., by interacting with a machine learning engine 128 to cause implementation of the model by the engine). In addition, the platform 122 can cache the content (e.g., within the repository 130) so that it can be used for subsequent prediction requests; para. [0063]) . Sanketi does not explicitly teach partitioning, the set of users into a first group of users and a second group of users; generating a first training dataset for the first group of users; generating a second training dataset for the second group of users; training, using the first training dataset and the second training dataset, a machine learning model configured to predict, the training comprising applying differential privacy to the second training dataset without applying differential privacy to the first training dataset. However, Amid teaches partitioning, for the subset of users, the set of users into a first group of users and a second group of users (i.e. We consider the DP-SCO setting with heterogeneous data, where there are two datasets Dpriv (with npriv samples) and Dpub (with npub samples) drawn i.i.d. from the same distribution. The private dataset Dpriv requires privacy protection, whereas the public dataset Dpub does not; Section 1.1, 5.2, pages 3, 11, 12) ; generating a first training dataset based on the user data for the first group of users; generating a second training dataset based on the user data for the second group of users (i.e. We consider the DP-SCO setting with heterogeneous data, where there are two datasets Dpriv (with npriv samples) and Dpub (with npub samples) drawn i.i.d. from the same distribution. The private dataset Dpriv requires privacy protection, whereas the public dataset Dpub does not; Section 1.1, 5.2, pages 3, 11, 12) ; training, using the first training dataset and the second training dataset, a machine learning model configured to predict information about one or more users (i.e. training, using the first training dataset and the second training dataset, a machine learning model configured to predict information about one or more users, the training comprising applying differential privacy to the second training dataset without applying differential privacy to the first training dataset; Section 1, 1.2, 5.2) , the training comprising applying differential privacy to the second training dataset without applying differential privacy to the first training dataset (i.e. We consider the DP-SCO setting with heterogeneous data, where there are two datasets Dpriv (with npriv samples) and Dpub (with npub samples) drawn i.i.d. from the same distribution. The private dataset Dpriv requires privacy protection, whereas the public dataset Dpub does not; Section 1.1, 3.1, pages 3, 11, 12) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the invention of Sanketi to include the feature of Amid. One would have been motivated to make this modification because it allows privacy restricted user data to contribute to training while preserving privacy, and to improve the privacy tradeoff by using non-private data to assist DP model training. Claim 2: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi further teaches wherein the machine learning model is configured to (i) process an input (i.e. the applications can communicate with the on-device machine learning platform via an API (which may be referred to as the “prediction API”) to provide input data and obtain predictions based on the input data from one or more of the machine-learned models; para. [0030, 0118]) comprising data specifying one or more contextual signals (i.e. the context features can be grouped or otherwise categorized according to a number of different context types. In general, each context type can specify or include a set of context features with well-known names and well-known types. One example context type is device information which includes the following example context features: audio state, network state, power connection; para. [0038, 0070]) included in a digital component request from a client device (i.e. when a particular application or other client requests (e.g., via the prediction API) for an inference to be generated on the basis of some client-provided input data, the context provider can inject or provide supplemental context features for input into the corresponding machine-learned model alongside the input data; para. [0038]) and to (ii) generate prediction data about the user of the client device based on the input (i.e. Thus, inferences can be made based at least in part on context information in addition to the client-provided input data, which may assist in improving the accuracy of the inferences; para. [0038]) . Claim 5: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi does not explicitly teach wherein training the machine learning model on the second training dataset with differential-privacy processing comprises: applying differentially private stochastic gradient descent (DP-SGD) to update model parameters of the machine learning model using the second training dataset. However, Amid further teaches wherein training the machine learning model on the second training dataset with differential-privacy processing (i.e. The private dataset Dpriv requires privacy protection, whereas the public dataset Dpub does not; Section 1.1) comprises: applying differentially private stochastic gradient descent (DP-SGD) (i.e. Gradient Descent (DP-SGD) [1, 6, 41], and its variants [26] have become the de facto standard algorithms for training machine learning models with differential privacy (DP); Section 1) to update model parameters of the machine learning model using the second training dataset (Algorithm 1, Public Data-Assisted Differentially Private Mirror Descent; page 4) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the invention of Sanketi to include the feature of Amid. One would have been motivated to make this modification because it allows privacy restricted user data to contribute to training while preserving privacy, and to improve the privacy tradeoff by using non-private data to assist DP model training. Claim 11: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi further teaches wherein the trained machine learning model is configured to output (i.e. The prediction API 412 can allow a client to feed input and derive predictions from it based on a trained model. Like the trainer, the predictor can be plan-driven in some implementations, where the plan is a declarative description of what operations to perform on a graph, and how to obtain inputs and produce outputs; para. [0118, 0121]) , based on the contextual signals in the digital component request, predicted data about the user of the digital component request (i.e. when a particular application 120 a-c or other client requests (e.g., via the prediction API) for an inference to be generated on the basis of some client-provided input data, the context provider can inject or provide supplemental context features for input into the corresponding machine-learned model 132 a-c alongside the input data. Thus, inferences can be made based at least in part on context information in addition to the client-provided input data, which may assist in improving the accuracy of the inferences; para. [0072, 0135]) . Claim 16: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi further teaches wherein the respective set of user data comprises user profile data (i.e. Example context features include: Audio State; Day Attributes; Calendar; Detected Activity; User-Specific Places (e.g., “home” vs. “work”; Network State; Power Connection; Screen Features; User Location; User Location Forecast; WiFi Scan Info; Weather; or other context features; para. [0135]) for a user registered on a digital service platform (i.e. Performance of the on-device machine learning functions on behalf of the one or more locally-stored applications or routines (which may be referred to as “clients”) may be provided as a centralized service to those clients, which may interact with the on-device machine learning platform via one or more application programming interfaces (APIs); para. [0027, 0031]) . Claim 17 is similar in scope to Claim 1 and is rejected under a similar rationale. Sanketi teaches a system comprising: one or more computers; and one or more storage devices storing instructions that when executed by the one or more computers, cause the one or more computers to perform operations comprising (i.e. The computing device 102 includes one or more processors 112 and a memory 114. The one or more processors 112 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, a FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. The memory 114 can include one or more non-transitory computer-readable storage mediums, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 114 can store data and instructions which are executed by the processor 112 to cause the computing device 102 to perform operations. The computing device 102 can also include a network interface 116 that enables communications over one or more networks (e.g., the Internet); para. [0058]) . Claim 18 is similar in scope to Claim 1 and is rejected under a similar rationale. Sanketi teaches one or more computer-readable storage media storing instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising (i.e. the computing device includes one or more processors and one or more non-transitory computer-readable media. The one or more non-transitory computer-readable media store: one or more applications implemented by the one or more processors; a centralized example database that stores training examples received from the one or more applications; para. [0010]) . Claim 19 is similar in scope to Claim 5 and is rejected under a similar rationale . 07-21-aia AIA 7. Claim s 3 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Sanketi in view of Amid, and further in view of Singh et al. (U.S. Patent Application Pub. No. US 20210201349 A1) Claim 3: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi further teaches wherein the machine learning model is configured to (i) process an input (i.e. The prediction API 412 can allow a client to feed input and derive predictions from it based on a trained model; para. [0118]) comprising data characterizing a digital component (i.e. They can also specify which context features should be automatically injected into candidate examples before being passed into the prediction engine; para. [0119, 0120]) to (ii) generate prediction data about for the digital component based on the input (i.e. The prediction API 412 can allow a client to feed input and derive predictions from it based on a trained model; para. [0118]) . Sanketi does not explicitly teach generate prediction data about an audience segment for the digital component based on the input. However, Singh teaches wherein the machine learning model is configured to (i) process an input comprising data characterizing a digital component (i.e. Once extracted, these elements are computationally transformed into numeric feature vectors that can be projected into a multidimensional space and used for downstream modeling as shown in FIGS. 5 and 6. FIG. 5 shows how supervised machine learning is used to build a regression model to predict a video performance score (computed from engagement metrics including views, likes, dislikes, upvotes, downvotes, shares, links, quotes, and comments) from the vectorized, extracted video content features; para. [0040]) to (ii) generate prediction data about an audience segment for the digital component based on the input (i.e. FIG. 6 shows how supervised machine learning is used to build a classification model to predict the proportion of audience engagement by demographics (computed from the natural language classifier pipeline shown in FIGS. 1-3) from the vectorized, extracted video content features; para. [0037, 0040]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Singh. One would have been motivated to make this modification because using content features and audience data to improve audience targeting, content recommendation, advertising, and marketing effectiveness. Claim 15: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi further teaches wherein the trained machine learning model is configured to output (i.e. The prediction API 412 can allow a client to feed input and derive predictions from it based on a trained model; para. [0118]) , based on the data characterizing the digital component (i.e. They can also specify which context features should be automatically injected into candidate examples before being passed into the prediction engine; para. [0119, 0120]) , predicted data for the digital component (i.e. The prediction API 412 can allow a client to feed input and derive predictions from it based on a trained model; para. [0118]) . Sanketi does not explicitly teach predicted data about one or more audience segments for the digital component. However, Singh teaches wherein the trained machine learning model is configured to output, based on the data characterizing the digital component (i.e. Once extracted, these elements are computationally transformed into numeric feature vectors that can be projected into a multidimensional space and used for downstream modeling as shown in FIGS. 5 and 6. FIG. 5 shows how supervised machine learning is used to build a regression model to predict a video performance score (computed from engagement metrics including views, likes, dislikes, upvotes, downvotes, shares, links, quotes, and comments) from the vectorized, extracted video content features; para. [0040]) , predicted data about one or more audience segments for the digital component (i.e. FIG. 6 shows how supervised machine learning is used to build a classification model to predict the proportion of audience engagement by demographics (computed from the natural language classifier pipeline shown in FIGS. 1-3) from the vectorized, extracted video content features; para. [0037, 0040]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Singh. One would have been motivated to make this modification because using content features and audience data to improve audience targeting, content recommendation, advertising, and marketing effectiveness . 07-21-aia AIA 8. Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Sanketi in view of Amid, and further in view of Shah et al. (U.S. Patent Application Pub. No. US 20120046996 A1) Claim 4: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi further teaches wherein the user data comprises (i) first user attribute data generated on a first content platform (i.e. the on-device machine learning platform can receive training examples from the applications via an API (which may be referred to as the “collection API”) and can manage storage of the examples in the centralized example database; para. [0032]) and (ii) first consent data controlling usage of the first user attribute data (i.e. the training examples and context features described herein are simply provided for the purposes of illustrating example data that could be stored with training examples or used to provide inferences by the on-device platform. However, such data is not collected, used, or analyzed unless the user has provided consent after being informed of what data is collected and how such data is used. Further, the user can be provided with a tool to revoke or modify the scope of permissions. In addition, certain information or data can be treated in or more ways before it is stored or used, so that personally identifiable information is removed or stored in an encrypted fashion; para. [0039]) . Sanketi does not explicitly teach one or more second content platforms. However, Shah teaches wherein the user data comprises (i) first user attribute data generated on a first content platform (i.e. The data provider sends user data to the DMP; para. [0077, 0089]) and (ii) first consent data controlling usage of the first user attribute data (i.e. controlling data permissions by client, tracking data utilization, and attributing and reporting data cost; para. [0008, 0066]) on one or more second content platforms (i.e. The platform provides solutions that address how to leverage custom audience segments across multiple demand side platforms (DSPs) and multiple media channels, such as display, video, mobile, digital TV, and digital-out-of-home, and provides approaches that allow management of all aspects of Internet advertising from a custom domain; abs) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Shah. One would have been motivated to make this modification because it improves cross-platform advertising/content targeting while maintaining client/data-source permission restrictions . 07-21-aia AIA 9. Claim s 6-9 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Sanketi in view of Amid, and further in view of Srinivasaraghavan (U.S. Patent Application Pub. No. US 20230057423 A1) . Claim 6: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi further teaches wherein: the user data for each user of the set of users comprises location data indicating a geographic region of the user (i.e. Example context features include: Audio State; Day Attributes; Calendar; Detected Activity; User-Specific Places (e.g., “home” vs. “work”; Network State; Power Connection; Screen Features; User Location; User Location Forecast; WiFi Scan Info; Weather; or other context features; para. [0135]) . Sanketi does not explicitly teach partitioning the set of users into the first group and the second group is further based on the geographic region of the user. However, Srinivasaraghavan teaches wherein: the user data for each user of the set of users comprises location data indicating a geographic region of the user; and partitioning the set of users into the first group and the second group is further based on the geographic region of the user (i.e. a user cluster can be associated with a group of users that are interested in the same type of content and/or are front the same geographic region or demographic grouping, or other type of data that can be used to group similar users and/or similar activities of users, in some embodiments, such cluster can be represented by n-dimensional vector s that represent values of interest and/or affiliation of the user to each topic or context that each cluster represents; para. [0072]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Srinivasaraghavan. One would have been motivated to make this modification because it improves the relevance and accuracy of content personalization and recommendation models. Claim 7: Sanketi, Amid, and Srinivasaraghavan teach the computer-implemented method of claim 6. Sanketi further teaches determining, based on the location data, that a user is located in a first geographic region (i.e. user's identity may be treated so that no personally identifiable information can be determined for the user, or a user's geographic location may be generalized where location information is obtained (such as to a city, ZIP code, or state level), so that a particular location of a user cannot be determined. Thus, the user may have control over what information is collected about the user, how that information is used, and what information is provided to the user; para. [0135, 0136]) ; assigning the user to the group if consent data is available for the user and the consent data for the user indicates the user permitting the one or more uses of the set of user data, and assigning the user to the group if consent data is unavailable for the user or the consent data for the user does not indicate the user permitting the one or more uses of the set of user data (i.e. the training examples and context features described herein are simply provided for the purposes of illustrating example data that could be stored with training examples or used to provide inferences by the on-device platform. However, such data is not collected, used, or analyzed unless the user has provided consent after being informed of what data is collected and how such data is used. Further, the user can be provided with a tool to revoke or modify the scope of permissions. In addition, certain information or data can be treated in or more ways before it is stored or used, so that personally identifiable information is removed or stored in an encrypted fashion; para. [0039, 0040]) . Sanketi does not explicitly teach wherein partitioning the set of users into the first group and the second group comprises: and in response to determining that the user is located in the first geographic region, assigning the user to the first group, and assigning the user to the second group. However, Srinivasaraghavan further teaches wherein partitioning the set of users into the first group and the second group comprises: determining, based on the location data, that a user is located in a first geographic region; and in response to determining that the user is located in the first geographic region, assigning the user to the first group, and assigning the user to the second group (i.e. a user cluster can be associated with a group of users that are interested in the same type of content and/or are front the same geographic region or demographic grouping, or other type of data that can be used to group similar users and/or similar activities of users, in some embodiments, such cluster can be represented by n-dimensional vector s that represent values of interest and/or affiliation of the user to each topic or context that each cluster represents; para. [0072, 0074]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Srinivasaraghavan. One would have been motivated to make this modification because it improves the relevance and accuracy of content personalization and recommendation models. Claim 8: Sanketi, Amid, and Srinivasaraghavan teach the computer-implemented method of claim 6. Sanketi further teaches determining, based on the location data, that a user is located in a second geographic region (i.e. user's identity may be treated so that no personally identifiable information can be determined for the user, or a user's geographic location may be generalized where location information is obtained (such as to a city, ZIP code, or state level), so that a particular location of a user cannot be determined. Thus, the user may have control over what information is collected about the user, how that information is used, and what information is provided to the user; para. [0135, 0136]) . Sanketi does not explicitly teach wherein partitioning the set of users into the first group and the second group comprises: in response to determining that the user is located in the second geographic region, assigning the user to the first group. However, Srinivasaraghavan further teaches wherein partitioning the set of users into the first group and the second group comprises: determining, based on the location data, that a user is located in a second geographic region; and in response to determining that the user is located in the second geographic region, assigning the user to the first group (i.e. a user cluster can be associated with a group of users that are interested in the same type of content and/or are front the same geographic region or demographic grouping, or other type of data that can be used to group similar users and/or similar activities of users, in some embodiments, such cluster can be represented by n-dimensional vector s that represent values of interest and/or affiliation of the user to each topic or context that each cluster represents; para. [0072, 0074]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Srinivasaraghavan. One would have been motivated to make this modification because it improves the relevance and accuracy of content personalization and recommendation models. Claim 9: Sanketi, Amid, and Srinivasaraghavan teach the computer-implemented method of claim 6. Sanketi further teaches determining, based on the location data, that a user is located in a third geographic region (i.e. user's identity may be treated so that no personally identifiable information can be determined for the user, or a user's geographic location may be generalized where location information is obtained (such as to a city, ZIP code, or state level), so that a particular location of a user cannot be determined. Thus, the user may have control over what information is collected about the user, how that information is used, and what information is provided to the user; para. [0135, 0136]) . Sanketi does not explicitly teach wherein partitioning the set of users into the first group and the second group comprises: in response to determining that the user is located in the third geographic region, assigning the user to the second group. However, Srinivasaraghavan further teaches wherein partitioning the set of users into the first group and the second group comprises: determining, based on the location data, that a user is located in a third geographic region; and in response to determining that the user is located in the third geographic region, assigning the user to the second group (i.e. a user cluster can be associated with a group of users that are interested in the same type of content and/or are front the same geographic region or demographic grouping, or other type of data that can be used to group similar users and/or similar activities of users, in some embodiments, such cluster can be represented by n-dimensional vector s that represent values of interest and/or affiliation of the user to each topic or context that each cluster represents; para. [0072, 0074]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Srinivasaraghavan. One would have been motivated to make this modification because it improves the relevance and accuracy of content personalization and recommendation models. Claim 20 is similar in scope to Claim 6 and is rejected under a similar rationale . 07-21-aia AIA 10. Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Sanketi in view of Amid, Srinivasaraghavan, and further in view of Tawakol et al. (U.S. Patent Application Pub. No. US 20160142379 A1) . Claim 10: Sanketi, Amid, and Srinivasaraghavan teach the computer-implemented method of claim 6. Sanketi further teaches determining, based on the location data, that a user is located in a fourth geographic region (i.e. user's identity may be treated so that no personally identifiable information can be determined for the user, or a user's geographic location may be generalized where location information is obtained (such as to a city, ZIP code, or state level), so that a particular location of a user cannot be determined. Thus, the user may have control over what information is collected about the user, how that information is used, and what information is provided to the user; para. [0135, 0136]) . Sanketi does not explicitly teach wherein partitioning the set of users into the first group and the second group comprises, for each user: determining, based on the location data, that a user is located in a fourth geographic region; and in response to determining that the user is located in the fourth geographic region, excluding the user from both the first and the second group. However, Srinivasaraghavan further teaches wherein partitioning the set of users into the first group and the second group comprises, for each user: determining, based on the location data, that a user is located in a fourth geographic region (i.e. a user cluster can be associated with a group of users that are interested in the same type of content and/or are front the same geographic region or demographic grouping, or other type of data that can be used to group similar users and/or similar activities of users, in some embodiments, such cluster can be represented by n-dimensional vector s that represent values of interest and/or affiliation of the user to each topic or context that each cluster represents; para. [0072, 0074]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Srinivasaraghavan. One would have been motivated to make this modification because it improves the relevance and accuracy of content personalization and recommendation models. However, Tawakol teaches determining, based on the location data, that a user is located in a fourth geographic region; and in response to determining that the user is located in the fourth geographic region, excluding the user from both the first and the second group (i.e. Strictly as some examples, an advertiser might be interested in presenting advertisements to a group of users whom are interested in “auto”. Using a retrieval mechanism such as a query to a database, the set of all users who have expressed interest in “auto” can be a starting point, however, to satisfy the aforementioned aspect of having only users in the group that do not have mutually-exclusive characteristics, some users might be rejected out. Some examples of mutually-exclusive characteristics include: (a) gender, (b) income bracket, (c) marital status, etc. When delivering groups of users that are similar, some users might be rejected out due to the presence of mutually-exclusive attributes, and even when a group of users do not have any mutually-exclusive characteristics, the group makeup can be made even more similar by selecting-in users that share still more characteristics in common (e.g., geographic region). The present disclosure shows and discusses how to use a rulebase to reject-out some users from a group, and further discusses how to use a rulebase to select-in some users; para. [0019]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi, Amid, and Srinivasaraghavan to include the feature of Tawakol. One would have been motivated to make this modification because it improves the relevance and accuracy of content personalization and recommendation models . 07-21-aia AIA 11. Claim s 12 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Sanketi in view of Amid, and further in view of Liu et al. (U.S. Patent Application Pub. No. US 20200104340 A1) . Claim 12: Sanketi and Amid teach the computer-implemented method of claim 1. Sanketi does not explicitly teach wherein: generating the first training dataset comprises, for each of a set of aggregation keys, generating an aggregated data profile by aggregating the user data of a respective subset of the first group of users having electronic resource views that match the aggregation key; and generating the second training dataset comprises, for each of the set of aggregation keys, generating an aggregated data profile by aggregating the user data of a respective subset of the second group of users having electronic resource views that match the aggregation key. However, Liu teaches wherein: generating the first training dataset comprises, for each of a set of aggregation keys (i.e. Keys to the hash table may include different combinations of test attributes 210 (e.g., test keys, treatment assignments, etc.) and/or dimensions 208 (e.g., user countries, languages, industries, etc.) to be analyzed using A/B test 212. Each key may map to a hash bucket containing a bitmap that encodes user IDs 216 of all users that have the corresponding attributes (e.g., users that have the same treatment assignment in a given A/B test, users that are from the same country and have the same treatment assignment in an A/B test, etc.); para. [0041]) , generating an aggregated data profile by aggregating the user data (i.e. Aggregation apparatus 202 then uses bitmaps 228 and/or the corresponding hash tables to efficiently join and/or aggregate records containing the metrics into a number of histograms 220. For example, aggregation apparatus 202 may iterate through records containing user IDs 216, page load times experienced by the users, and/or page IDs and/or products associated with the page load times. As mentioned above, the records may be distributed across partitions by user IDs 216; para. [0042]) of a respective subset of the first group of users having electronic resource views that match the aggregation key (i.e. Each key may map to a hash bucket containing a bitmap that encodes user IDs 216 of all users that have the corresponding attributes (e.g., users that have the same treatment assignment in a given A/B test, users that are from the same country and have the same treatment assignment in an A/B test, etc.); para. [0034, 0041]) ; and generating the second training dataset comprises, for each of the set of aggregation keys, generating an aggregated data profile by aggregating the user data of a respective subset of the second group of users having electronic resource views that match the aggregation key (i.e. Aggregation apparatus 202 aggregates metrics collected during an A/B test 212. For example, the metrics may include page views, page load times, latencies, error rates, session counts, session lengths, CTRs, conversion rates, and/or other performance metrics that are used to compare the treatment and control variants of A/B test 212; para. [0033]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Liu. One would have been motivated to make this modification because it provides efficiently aggregating user page view records using aggregation keys. Claim 13: Sanketi, Amid, and Liu teach the computer-implemented method of claim 12. Sanketi does not explicitly teach for each of the set of aggregation keys, before aggregating the user data of the respective subset of the second group of users, applying differential privacy to the user data of the respective subset of the second group of users. However, Amid further teaches generating the second training dataset further comprises, applying differential privacy to the user data of the respective subset of the second group of users (i.e. We consider the DP-SCO setting with heterogeneous data, where there are two datasets Dpriv (with npriv samples) and Dpub (with npub samples) drawn i.i.d. from the same distribution. The private dataset Dpriv requires privacy protection, whereas the public dataset Dpub does not; Section 1.1, 5.2, pages 3, 11, 12) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the invention of Sanketi to include the feature of Amid. One would have been motivated to make this modification because it allows privacy restricted user data to contribute to training while preserving privacy, and to improve the privacy tradeoff by using non-private data to assist DP model training. However, Liu further teaches for each of the set of aggregation keys, before aggregating the user data (i.e. Keys to the hash table may include different combinations of test attributes 210 (e.g., test keys, treatment assignments, etc.) and/or dimensions 208 (e.g., user countries, languages, industries, etc.) to be analyzed using A/B test 212. Each key may map to a hash bucket containing a bitmap that encodes user IDs 216 of all users that have the corresponding attributes (e.g., users that have the same treatment assignment in a given A/B test, users that are from the same country and have the same treatment assignment in an A/B test, etc.); para. [0041]) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi and Amid to include the feature of Liu. One would have been motivated to make this modification because it provides efficiently aggregating user page view records using aggregation keys . 07-21-aia AIA 12. Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over Sanketi in view of Amid, and further in view of Steiman et al. (U.S. Patent Application Pub. No. US 20220036538 A1) . Claim 14: Sanketi, Amid, and Liu teach the computer-implemented method of claim 12. Sanketi does not explicitly teach wherein training the machine learning model comprises adding the aggregated data profiles as training labels of the machine learning model. However, Steiman teaches wherein training the machine learning model comprises adding the aggregated data profiles as training labels of the machine learning model (i.e. [0009] The criterion is based on a user feedback on the first segmentation map, and the PMC is configured to, upon receiving a negative user feedback on the first segmentation map, obtain additional first labels associated with an additional group of pixels in at least one of the segments, the first labels and the additional first labels constituting aggregated label data, and repeat the extracting of the second features, the training and the processing based on the aggregated label data until receiving a positive user feedback. [0010] (iii). The PMC is configured to, upon receiving a positive user feedback on the first segmentation map, include the first training sample into the training data. [0011] (iv). The PMC is further configured to obtain a second training image and second labels respectively associated with a group of pixels selected in each of one or more segments identified by a user from the second training image, the second labels being added to the aggregated label data, extract a set of features characterizing the second training image and including the first features and the second features, train the ML model using the aggregated label data, values of pixels associated with the aggregated label data, and the feature values of each feature of the set of features corresponding to the pixels associated with the aggregated label data, and performing the processing and determining based on the second training image) . Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Sanketi, Amid, and Liu to include the feature of Steiman. One would have been motivated to make this modification because it improves model training and prediction quality . Conclusion 07-96 AIA The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure. Carley (Pub. No. US 20220067181 A1), The data being classified or used for training machine learning models is often sensitive and may come from multiple sources with different privacy requirements. Such data may also have value in view of the investment to acquire the datasets, perform preliminary processing, perform labeling, and/or store the dataset. In fact, many datasets (e.g., those used for controlling autonomous vehicles) are built after laborious processes that merge and clean multiple sources of data, and annotate the datasets with appropriate labels (often proprietary). Therefore, datasets are also based on proprietary or private data that owners do not want to share. It is noted that any citation to specific pages, columns, lines, or figures in the prior art references and any interpretation of the references should not be considered to be limiting in any way. A reference is relevant for all it contains and may be relied upon for all that it would have reasonably suggested to one having ordinary skill in the art. In re Heck , 699 F.2d 1331, 1332-33, 216 U.S.P.Q. 1038, 1039 (Fed. Cir. 1983) (quoting In re Lemelson , 397 F.2d 1006, 1009, 158 U.S.P.Q. 275, 277 (C.C.P.A. 1968)) . Any inquiry concerning this communication or earlier communications from the examiner should be directed to TAN TRAN whose telephone number is (303)297-4266. The examiner can normally be reached on Monday - Thursday - 8:00 am - 5:00 pm MT. 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, Matt Ell can be reached on 571-270-3264. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300. Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000. /TAN H TRAN/Primary Examiner, Art Unit 2141 Application/Control Number: 18/574,668 Page 2 Art Unit: 2141 Application/Control Number: 18/574,668 Page 3 Art Unit: 2141 Application/Control Number: 18/574,668 Page 4 Art Unit: 2141 Application/Control Number: 18/574,668 Page 5 Art Unit: 2141 Application/Control Number: 18/574,668 Page 6 Art Unit: 2141 Application/Control Number: 18/574,668 Page 7 Art Unit: 2141 Application/Control Number: 18/574,668 Page 8 Art Unit: 2141 Application/Control Number: 18/574,668 Page 9 Art Unit: 2141 Application/Control Number: 18/574,668 Page 10 Art Unit: 2141 Application/Control Number: 18/574,668 Page 11 Art Unit: 2141 Application/Control Number: 18/574,668 Page 12 Art Unit: 2141 Application/Control Number: 18/574,668 Page 13 Art Unit: 2141 Application/Control Number: 18/574,668 Page 14 Art Unit: 2141 Application/Control Number: 18/574,668 Page 15 Art Unit: 2141 Application/Control Number: 18/574,668 Page 16 Art Unit: 2141 Application/Control Number: 18/574,668 Page 17 Art Unit: 2141 Application/Control Number: 18/574,668 Page 18 Art Unit: 2141 Application/Control Number: 18/574,668 Page 19 Art Unit: 2141 Application/Control Number: 18/574,668 Page 20 Art Unit: 2141 Application/Control Number: 18/574,668 Page 21 Art Unit: 2141 Application/Control Number: 18/574,668 Page 22 Art Unit: 2141 Application/Control Number: 18/574,668 Page 23 Art Unit: 2141 Application/Control Number: 18/574,668 Page 24 Art Unit: 2141 Application/Control Number: 18/574,668 Page 25 Art Unit: 2141 Application/Control Number: 18/574,668 Page 26 Art Unit: 2141
Prosecution Insights