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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114 was filed in this application after a decision by the Patent Trial and Appeal Board, but before the filing of a Notice of Appeal to the Court of Appeals for the Federal Circuit or the commencement of a civil action. Since this application is eligible for continued examination under 37 CFR 1.114 and the fee set forth in 37 CFR 1.17(e) has been timely paid, the appeal has been withdrawn pursuant to 37 CFR 1.114 and prosecution in this application has been reopened pursuant to 37 CFR 1.114. Applicant’s submission filed on July 7, 2026 has been entered.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on July 14, 2026 is 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 § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-4, 6, 10-13, 15-18 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Tasinga et al. (U.S. Patent Publication No. 2022/0180178) (“Tasinga”) in view of Gubin et al. (U.S. Patent Publication No. 2013/0031489) (“Gubin”) in further view of Gopalan (U.S. Patent Publication No. 2018/0285772) (“Gopalan”).
Regarding claim 1, Tasinga teaches a system comprising: at least one hardware processor (Tasinga [0060] “In at least one embodiment, servers 110-114 each comprise at least one processor and at least one memory for storing processor-executable instructions.” Tasinga provides a system comprising at least one processor.); and a computer-readable medium storing instructions that, when executed by the at least one hardware processor, cause the at least one hardware processor to perform operations (Tasinga [0060] “In at least one embodiment, servers 110-114 each comprise at least one processor and at least one memory for storing processor-executable instructions.” Tasinga provides a memory for storing processor-executable instructions corresponding to a computer-readable medium storing instructions that, when executed by the at least one hardware processor, cause the at least one hardware processor to perform operations.) comprising: obtaining an intelligent scheduling machine learned model trained using a first machine learning algorithm (Tasinga [0065] “In at least one embodiment, an AI-assisted scheduler 102 maintains one or more neural networks trained to predict performance characteristics of machine learning models 120” Tasinga provides an AI assisted scheduler trained to predict model performance characteristics of other models corresponding to obtaining an intelligent scheduling machine learned model trained using a first machine learning algorithm.), the training comprising obtaining a first set of training data and passing the first set of training data through the first machine learning algorithm (Tasinga [0065] “In at least one embodiment, one or more neural networks are trained to predict performance of a model M.sub.1 on a particular server S.sub.1, in view of past performance characteristics of said model M.sub.1 on said server S.sub.1. In at least one embodiment, input to said one or more neural networks comprises an indication of an identity of model M.sub.1 and an indication of an identity server S.sub.1.” Tasinga provides training comprising obtaining past performance characteristics corresponds to training data which is passed through the machine learning algorithm since it occurs during the training.) to learn a coefficient for each of a plurality of features of the training data (Tasinga [0060] “In at least one embodiment, servers 110-114 each comprise or are connected to a storage device containing data pertaining to a machine learning model, such as weights, biases, parameters, and so forth.” Tasinga provides data used for training comprising weights and model parameters, which corresponds to the scheduler learning coefficients for each of a plurality of features of training data.) …receiving, at an application server in a cloud environment, a request (Tasinga [0296] “In at least one embodiment, a work descriptor (WD) 1784 contained in process element 1783 can be a single job requested by an application or may contain a pointer to a queue of jobs.”; [0526] “In at least one embodiment, object storage may be accessible through, for example, a cloud storage (e.g., a cloud 3926 of FIG. 39) compatible application programming interface (API) from within a cloud platform.” Tasinga provides receiving requests at an application server in a cloud environment.) to generate a first inference model for a first entity of a plurality of entities corresponding to the cloud environment (Tasinga [0532] “In at least one embodiment, a data processing pipeline may receive input data (e.g., imaging data 3808) in a DICOM, RIS, CIS, REST compliant, RPC, raw, and/or other format in response to an inference request (e.g., a request from a user of deployment system 3806, such as a clinician, a doctor, a radiologist, etc.).” Tasinga provides receiving inference requests corresponding to a request to generate a first inference model for clinicians, doctors, radiologists, etc., that corresponds to a first entity for a plurality if entities, which are described in specification paragraph [0040] as groups of users.); in response to the receiving, causing a first version of the first inference model to be generated (Tasinga [0525] “In at least one embodiment, training system 3804 may be used to provide applications, services, and/or other resources for generating working, deployable machine learning models for deployment system 3806.” Tasinga provides generating models which includes models used for inferencing as described above.) and trained using a second machine learning algorithm (Tasinga [0528] “In at least one embodiment, once a model is trained—or partially trained—at one location, a machine learning model may be added to model registry 3824. In at least one embodiment, a machine learning model may then be retrained, or updated, at any number of other facilities, and a retrained or updated model may be made available in model registry 3824. In at least one embodiment, a machine learning model may then be selected from model registry 3824—and referred to as output model 3816—and may be used in deployment system 3806 to perform one or more processing tasks for one or more applications of a deployment system.” Tasinga provides updating already trained models, which corresponds to using second ML algorithms to train a first model being generated.) and a second set of training data (Tasinga [0525] “In at least one embodiment, machine learning models may be trained at facility 3802 using data 3808 (such as imaging data) generated at facility 3802 (and stored on one or more picture archiving and communication system (PACS) servers at facility 3802), may be trained using imaging or sequencing data 3808 from another facility or facilities (e.g., a different hospital, lab, clinic, etc.), or a combination thereof.” Tasinga provides training inference models using a variety of training data sets from different sources corresponding to trained using a second set of data.)…
Tasinga fails to teach … the intelligent scheduling machine learned model being trained to dynamically learn a different retraining frequency for each combination of a plurality of combinations of entities and inference models by applying the learned coefficient for each of the plurality of features to a value for that feature … causing retraining of the first inference model at an initial retraining frequency; determining that a change has occurred in how often new training data is received; in response to the determination, inputting a set of features corresponding to the first entity and to the first inference model to the intelligent scheduling machine learned model to generate a first retraining frequency for the first inference model, the set of features including a value indicating an amount of change in a distribution pattern of the training data for each of a plurality of labels, the amount of change being determined by random sampling of the training data since a prior training or retraining of the first inference model; causing retraining of a first version of the first inference model, using the new training data, at the first retraining frequency instead of the initial retraining frequency; inputting a set of features corresponding to a second entity and to the first inference model to the intelligent scheduling machine learned model to generate a second retraining frequency for the first inference model; causing retraining of a second version of the first inference model, different than the first version, at the second retraining frequency.
However, Gubin teaches …the intelligent scheduling machine learned model being trained to dynamically learn a different retraining frequency for each combination of a plurality of combinations of entities and inference models (Gubin [0039] “The machine learning module 345 uses machine learning techniques to generate the news feed ranking model 125. In an embodiment, the machine learning module 345 may generate a portion of the functionality invoked by the news feed ranking model 125. For example, the machine learning module 345 may generate a model that determines a ranking score associated with a given news feed story 180. The news feed ranking model 125 can order a set of news feed stories based on their ranking scores.”; [0072] “The model corresponding to each subset may be retrained 630 at a different frequency compared to other models. The machine learning module 345 may use information describing how users of each subset interact with the news feed stories ranked by a model are presented to them to determine a rate at which the corresponding model is retrained 630.” Gubin provides a machine learning module 345 which uses machine learning techniques to determine a rate at which models are retrained, wherein each model is associated with a subset of users (entity), corresponding to a machine learned model being trained to dynamically learn a different retraining frequency for each combination of a plurality of combinations of entities and inference models.) by applying the learned coefficient for each of the plurality of features to a value for that feature (Gubin [0054] “The machine learning module 345 may retrieve the features from the feature store 325 if the features have been previously extracted by the feature extraction module 310 and stored in the feature store 325. The features extracted 445 may be associated with individual objects or with interactions between objects.”; [0071] “The generation 620 of the news feed ranking model 125 may comprise training the news feed ranking model 125 based on a training data set. The training data set may be provided by an expert. An expert may analyze various features of the news feed stories 180 and determines data corresponding to values of the feature” Gubin teaches training the model using features with corresponding values for each feature, corresponding to applying a learned coefficient (i.e., during training) for each of a plurality of features to a value for that feature.) …causing retraining of the first inference model at an initial retraining frequency (Gubin [0049] “If there is more than a threshold change in the access rate compared to an expected access rate, the news feed access analyzer module 315 may indicate to the machine learning module 345 to retrain the news feed ranking model 125 for the corresponding group of users.”; [0072] “The machine learning module 345 periodically retrains 630 the model corresponding to each subset of users of the social networking system.” Gubin provides periodic retraining of machine learning models corresponding to causing retraining of the first inference model at an initial retraining frequency.); …in response to the determination, inputting a set of features corresponding to the first entity and to the first inference model to the intelligent scheduling machine learned model to generate a first retraining frequency for the first inference model (Gubin [0072] “The model corresponding to each subset may be retrained 630 at a different frequency compared to other models. The machine learning module 345 may use information describing how users of each subset interact with the news feed stories ranked by a model are presented to them to determine a rate at which the corresponding model is retrained 630.” Gubin provides using information describing how users of each subset interact with the news feed stories ranked by a model are presented to them to determine a rate at which the corresponding model is retrained, corresponding in response to a determination of new data, inputting a set of features corresponding to the first entity and to the first inference model to the intelligent scheduling machine learned model to generate a first retraining frequency for the first inference model.) …causing retraining of a first version of the first inference model, using the new training data, at the first retraining frequency instead of the initial retraining frequency (Gubin [0072] “The machine learning module 345 may use information describing how users of each subset interact with the news feed stories ranked by a model are presented to them to determine a rate at which the corresponding model is retrained 630.”; [0073] “FIG. 7 is a flowchart of the process of periodically retraining a model for ranking news feed stories for a set of users of the social networking system, in accordance with one embodiment of the invention. The machine learning module 345 retrains the model 125 based on changes in access patterns of news feed stories 180 by the users. The news feed access analyzer 315 identifies 720 changes in access patterns of news feed stories by the users presented with the stories.” Gubin provides determining different retraining rates based on changing information corresponding to causing retraining of a first version of the first inference model at the first retraining frequency instead of the initial retraining frequency using new training data.); inputting a set of features corresponding to a second entity and to the first inference model to the intelligent scheduling machine learned model to generate a second retraining frequency for the first inference model (Gubin [0072] “The model corresponding to each subset may be retrained 630 at a different frequency compared to other models. The machine learning module 345 may use information describing how users of each subset interact with the news feed stories ranked by a model are presented to them to determine a rate at which the corresponding model is retrained 630.”; [0073] “The machine learning module 345 retrains the model 125 based on changes in access patterns of news feed stories 180 by the users. The news feed access analyzer 315 identifies 720 changes in access patterns of news feed stories by the users presented with the stories. The change in the access patterns of the news feed stories 180 can be caused by various factors. There may be an overall change in the demographics of the users, for example, the percentage of population with particular demographic characteristics that is member of the social networking system 100 may have increased.” Gubin provides inputting changes in access patterns of news feed stories 180 by the users to determine a rate at which at which a machine learning model is to be retrained associated with changing user demographics (a second entity), corresponding to inputting a set of features corresponding to a second entity and to the first inference model to the intelligent scheduling machine learned model to generate a second retraining frequency for the first inference model.); causing retraining of a second version of the first inference model, different than the first version, at the second retraining frequency (Gubin [0072] “The machine learning module 345 may use information describing how users of each subset interact with the news feed stories ranked by a model are presented to them to determine a rate at which the corresponding model is retrained 630.”; [0073] “FIG. 7 is a flowchart of the process of periodically retraining a model for ranking news feed stories for a set of users of the social networking system, in accordance with one embodiment of the invention. The machine learning module 345 retrains the model 125 based on changes in access patterns of news feed stories 180 by the users. The news feed access analyzer 315 identifies 720 changes in access patterns of news feed stories by the users presented with the stories.” Gubin provides determining different retraining rates based on changing information and changing user demographics corresponding to causing retraining of a second version of the first inference model, different than the first version, at the second retraining frequency.).
Tasinga and Gubin are both considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically machine learning model training/retraining scheduling. Therefore, it would have obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tasinga with the above teachings of Gubin. Doing so would allow to ensure that a model reflects the latest information available (Gubin [0017] “Each model may be retrained at a different rate to ensure that the model reflects the latest information available in the social networking system that affects the ranking of news feed stories.”).
Tasinga in view of Gubin fails to explicitly teach determining that a change has occurred in how often new training data is received; … the set of features including a value indicating an amount of change in a distribution pattern of the training data for each of a plurality of labels, the amount of change being determined by random sampling of the training data since a prior training or retraining of the first inference model.
However, Gopalan teaches determining that a change has occurred in how often new training data is received (Gopalan [0015] “In addition, in one example, one or more features with the greatest relevance score, and/or which have had a greatest increase in relevance score, may be identified for increased data collection or data sampling with respect to a new training data set (e.g., for further retraining of the machine learning model) and/or with respect to streams of new data to be processed by the retrained machine learning model. Alternatively, or in addition, one or more features with the lowest relevance scores, or with the greatest drops in relevance scores, may be designated for reduced data collection or data sampling with respect to a new training data set and/or with respect to streams of new data to be processed by the retrained machine learning model.” Gopalan teaches increased or decreased sampling of new training data, corresponding to a change in how often new training data is received.) …the set of features including a value indicating an amount of change in a distribution pattern of the training data for each of a plurality of labels (Gopalan [0031] “In another example, the training data set may include or comprise labeled data.”; [0033] “At step 230, the processor processes a stream of new data to determine a likelihood of the new data from the data distribution that is computed at step 220. The likelihood may comprise a Euclidean or non-Euclidean distance between the data distribution of the training data set and a data distribution of the stream of new data, e.g., with respect to a feature space comprising the same dimensions/feature vectors for both the data distribution of the training data set and the data distribution of the stream of new data. In another example, the likelihood may comprise an integral of the difference between the data distribution of the training data set and a data distribution of the stream of new data. In one example, the likelihood may be reported as a value (e.g., a percentage, or a score on a different scale).” Gopalan teaches a Euclidean distance value indicating an amount of change in a distribution of the training data, which comprises labeled data.), the amount of change being determined by random sampling of the training data since a prior training or retraining of the first inference model (Gopalan [0031] “In one example, a portion of the training data set is used to train the machine learning model. For instance, data associated with various features may be randomly sampled, sampled at regular data intervals, and so forth, such that less that all of the available training data set is used to train the machine learning model.” Gopalan teaches training a machine learning model including a random sampling of the training data at regular intervals, corresponding to a random sampling of training data since prior training.)
Tasinga, Gubin and Gopalan are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically machine learning model training/retraining scheduling. Therefore, it would have obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tasinga in view of Gubin with the above teachings of Gopalan. Doing so may improve the accuracy of a machine learning model (Gopalan [0048] “Furthermore, the telecommunication service provider network is also transformed via the automatic adjustment of data collection or data sampling rates to improve the accuracy of a machine learning model, in accordance with examples of the present disclosure.”).
Regarding claim 2, Tasinga in view of Gubin in further view of Gopalan teaches the system of claim 1 as discussed above in the rejection of claim 1, wherein the intelligent scheduling machine learned model is a neural network (Tasinga [0065] “In at least one embodiment, an AI-assisted scheduler 102 maintains one or more neural networks trained to predict performance characteristics of machine learning models 120.” Tasinga provides an AI-assisted scheduler corresponding to the intelligent scheduling machine learned model as neural network.).
It would have been obvious to one of ordinary skill in the art before the filing date of the claimed invention to combine the teaching of Tasinga, Gubin and Gopalan for the same reasons disclosed in claim 1 above.
Regarding claim 3, Tasinga in view of Gubin in further view of Gopalan teaches the system of claim 1 as discussed above on the rejection of claim 1, wherein the first entity is a group of users (Tasinga [0532] “In at least one embodiment, a data processing pipeline may receive input data (e.g., imaging data 3808) in a DICOM, RIS, CIS, REST compliant, RPC, raw, and/or other format in response to an inference request (e.g., a request from a user of deployment system 3806, such as a clinician, a doctor, a radiologist, etc.).” Tasinga provides an entity that includes groups of users, such as clinicians, doctors, radiologists, etc.).
It would have been obvious to one of ordinary skill in the art before the filing date of the claimed invention to combine the teaching of Tasinga, Gubin and Gopalan for the same reasons disclosed in claim 1 above.
Regarding claim 4, Tasinga in view of Gubin in further view of Gopalan teaches the system of claim 1, as discussed above on the rejection of claim 1 wherein the operations further comprise: receiving, at the application server, training data parameters for the first inference model (Tasinga [0116] “In at least one embodiment, inference and/or training logic 915 may include, without limitation, code and/or data storage 901 to store forward and/or output weight and/or input/output data, and/or other parameters to configure neurons or layers of a neural network trained and/or used for inferencing in aspects of one or more embodiments.” Tasinga provides parameters received for training of models/inference models.) and wherein the causing the first version of the first inference model to be generated and trained further includes filtering the second set of training data based on the training data parameters (Tasinga [0532] “In at least one embodiment, data may undergo pre-processing as part of data processing pipeline to prepare data for processing by one or more applications.” Tasinga provides pre-processing of data, which corresponds to filtering the second set of training data based on the training data parameters.).
It would have been obvious to one of ordinary skill in the art before the filing date of the claimed invention to combine the teaching of Tasinga, Gubin and Gopalan for the same reasons disclosed in claim 1 above.
Regarding claim 6, Tasinga in view of Gubin in further view of Gopalan teaches the system of claim 1 as discussed above on the rejection of claim 1, wherein the set of features corresponding to the first entity and to the first inference model includes information about a type associated with the second machine learning algorithm (Tasinga [0545] “In at least one embodiment, output model(s) 3816 and/or pre-trained model(s) 3906 may include any types of machine learning models depending on implementation or embodiment. In at least one embodiment, and without limitation, machine learning models used by system 3900 may include machine learning model(s) using linear regression, logistic regression, decision trees, support vector machines (SVM), Naïve Bayes, k-nearest neighbor (Knn), K means clustering, random forest, dimensionality reduction algorithms, gradient boosting algorithms, neural networks (e.g., auto-encoders, convolutional, recurrent, perceptrons, Long/Short Term Memory (LSTM), Hopfield, Boltzmann, deep belief, deconvolutional, generative adversarial, liquid state machine, etc.), and/or other types of machine learning models.” Tasinga provides multiple types of machine learning algorithms for training corresponding information about a type associated with the second machine learning algorithm.).
It would have been obvious to one of ordinary skill in the art before the filing date of the claimed invention to combine the teaching of Tasinga, Gubin and Gopalan for the same reasons disclosed in claim 1 above.
Regarding claim 10, it is the method embodiment of claim 1 with similar limitations to claim 1 and is rejected using the same reasoning found in the rejection of claim 1 above.
Regarding claim 11, the rejection of claim 10 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin and Gopalan for the same reasons disclosed above in the rejection of claim 2.
Regarding claim 12, the rejection of claim 10 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin and Gopalan for the same reasons disclosed above in the rejection of claim 3.
Regarding claim 13, the rejection of claim 10 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin and Gopalan for the same reasons disclosed above in the rejection of claim 4.
Regarding claim 15, it is the computer hardware embodiment of claim 1 with similar limitations to claim 1 and is rejected using the same reasoning found in the rejection of claim 1 above. Further Tasinga teaches a non-transitory machine-readable medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations (Tasinga [0651] “In at least one embodiment, code (e.g., executable code or source code) is stored on a set of one or more non-transitory computer-readable storage media having stored thereon executable instructions (or other memory to store executable instructions) that, when executed (i.e., as a result of being executed) by one or more processors of a computer system, cause computer system to perform operations” Tasinga provides a non-transitory machine-readable medium storing instructions storing instructions configured to perform operations by the execution of a processor.).
It would have been obvious to one of ordinary skill in the art before the filing date of the claimed invention to combine the teaching of Tasinga, Gubin and Gopalan for the same reasons disclosed in claim 1 above.
Regarding claim 16, the rejection of claim 15 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin in further view of Gopalan for the same reasons disclosed above in the rejection of claim 2.
Regarding claim 17, the rejection of claim 15 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin in further view of Gopalan for the same reasons disclosed above in the rejection of claim 3.
Regarding claim 18, the rejection of claim 15 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin in further view of Gopalan for the same reasons disclosed above in the rejection of claim 4.
Regarding claim 20, the rejection of claim 15 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin in further view of Gopalan for the same reasons disclosed above in the rejection of claim 6.
Claims 5, 14, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Tasinga et al. (U.S. Patent Publication No. 2022/0180178) (“Tasinga”) in view of Gubin et al. (U.S. Patent Publication No. 2013/0031489) (“Gubin”) in further view of Gopalan (U.S. Patent Publication No. 2018/0285772 ) (“Gopalan”) and Noorizadeh et al. (U.S. Patent Publication No. 2022/0058512) (“Noorizadeh”).
Regarding claim 5, Tasinga in view of Gubin in further view of Gopalan teaches the system of claim 1 as discussed above in the rejection of claim 1, but fails to teach wherein the operations further comprise repeating the inputting and retraining for a subsequent version of the first inference model, causing a different retraining frequency to be output and used.
However, Noorizadeh teaches wherein the operations further comprise repeating the inputting and retraining for a subsequent version of the first inference model (Noorizadeh [0119] “Storing retrained models in central model repository 705 together with the applied training data may allow AI engine 607 to determine whether model 515 needs to be retrained again in response to future changes to the training data” Noorizadeh provides retraining models that need to be retrained again, which corresponds to repeating the inputting and retraining for a subsequent version of the first inference model), causing a different retraining frequency to be output and used (Noorizadeh [0056] “The AI engine may generate an alert indicating that models associated with the specific operational domain have been undergoing more frequent model training than models associated with other operational domains.” Noorizadeh provides outputting retraining frequencies as an alert, which vary and which corresponds to causing a different retraining frequency to be output and used.).
Tasinga, Gubin, Gopalan and Noorizadeh are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically machine learning. Therefore, it would have obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tasinga in view of Gubin in further view of Gopalan with the above teachings of Noorizadeh. Doing so would allow for efficient retraining of machine learning models based on the calculated need/frequency of retraining thus reducing resource consumption (Noorizadeh [0007] “Some machine learning models may need to be retrained frequently and others less frequently. Some models may require even more power consumption to train. As described herein, a MACHINE LEARNING MODEL TRAINING SYSTEM provides technical solutions for reducing power consumption and utilization of model training systems.”).
Regarding claim 14, the rejection of claim 10 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin in further view of Gopalan and Noorizadeh for the same reasons disclosed above in the rejection of claim 5.
Regarding claim 19, the rejection of claim 15 is incorporated herein. Further, the limitations in this claim are taught by Tasinga in view of Gubin in further view of Gopalan and Noorizadeh for the same reasons disclosed above in the rejection of claim 5.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Tasinga et al. (U.S. Patent Publication No. 2022/0180178) (“Tasinga”) in view of Gubin et al. (U.S. Patent Publication No. 2013/0031489) (“Gubin”) in further view of Gopalan (U.S. Patent Publication No. 2018/0285772 ) (“Gopalan”) and Srinivasan et al. (U.S. Patent Publication No. 2018/0300653) (“Srinivasan”).
Regarding claim 7, Tasinga in view of Gubin in further view of Gopalan teaches the system of claim 1 as discussed above in the rejection of claim 1, but fails to teach wherein the operations further comprise: dynamically assigning training of the first inference model to a first container, the first container containing a first combination of computing resources used to perform the training of the first inference model; and based on the determination that the change has occurred in how often new training data is received, dynamically reassigning the training of the first inference model to a second container, the second container containing a second combination, different than the first combination, of computing resources used to perform the training of the first inference model.
However, Srinivasan teaches dynamically assigning training of the first inference model to a first container, the first container containing a first combination of computing resources used to perform the training of the first inference model (Srinivasan [0018] “In some implementations, the distributed machine learning system utilizes machine learning containers (“ML containers” or “containers”) to perform training tasks and prediction tasks (which may also be referred to as “training jobs” and “prediction jobs”).”; [0020] “The DML system generates a training container by retrieving a container image that is configured with a particular model architecture and by mounting a volume of training data to the filesystem of the container image. The DML system assigns the training container to one or more processors, where the processors can include special processors such as graphics processor units (GPUs). Once the training container begins executing on the assigned processor(s), the container is a self-contained computer and outputs a machine learned model, which may be used in a serving container to issue predictions.” Srinivasan provides a training container for training machine learning models for predictions, wherein each container receives one or more assigned processors, corresponding to dynamically assigning training of the first inference model to a first container, the first container containing a first combination of computing resources used to perform the training of the first inference model.); and based on the determination that the change has occurred in how often new training data is received, dynamically reassigning the training of the first inference model to a second container, the second container containing a second combination, different than the first combination, of computing resources used to perform the training of the first inference model (Srinivasan [0028] “In other scenarios, the base model may be a previous version of the model to be trained, whereby the container is updating the model weights in the model using newly received training data.”; [0059] “In some scenarios, the container manager 218 may select a base model from a previous training session (e.g., in the case of updating a model).”; [0062] “Upon mounting a volume 234 of training data and initializing the model architecture, the container manager 218 can deploy the training container on the appropriate computing resource. For instance, the container manager 218 can run a training container for training an image classifier on two or more available GPUs 214. The training container executes the computer executable instructions contained in the model architecture, which define operations to be performed on the matrix given the training data. As a training container executes, the training container updates the model at each iteration.”; [0073] “The DML system may determine the resources required based on a number of factors, including the model architecture, the type of training data (e.g., images require more resources than text), and the size of a volume of training data.”; [0079] “The base model may be a previously learned version of a model, a generic model, or a randomly initialized model. In the case that the training job is directed to updating a previously learned model, the DML system may utilize the previously learned model, whereby the container operates on the previously learned model using newly received training data.” Srinivasan provides utilizing a previous model in a new training container and training with newly received training data and different assigned processors/computing resources (GPUs 214), corresponding to dynamically reassigning the training of the first inference model to a second container, the second container containing a second combination, different than the first combination, of computing resources used to perform the training of the first inference model based on the determination that the change has occurred in how often new training data is received.).
Tasinga, Gubin, Gopalan and Srinivasan are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically machine learning model training. Therefore, it would have obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tasinga in view of Gubin and Gopalan with the above teachings of Srinivasan. Doing so would allow for computing resources to be distributed as needed for model training (Srinivasan [0026] “The DML system 200 may include a scheduler (not shown in FIG. 1A) that assigns training and/or prediction tasks to various resources of the DML system 200. For instance, some training tasks may require a graphical processing unit (GPU) while other training tasks may have higher priority than other tasks. The scheduler maintains a queue of tasks and also monitors the availability of the computing resources of the DML system 200. The scheduler assigns tasks to available resources based on the priority of a respective task, the minimum requirements of the respective task, and the available resources at the time.”).
Claim 21 is rejected under 35 U.S.C. 103 as being unpatentable over Tasinga et al. (U.S. Patent Publication No. 2022/0180178) (“Tasinga”) in view of Gubin et al. (U.S. Patent Publication No. 2013/0031489) (“Gubin”) in further view of Gopalan (U.S. Patent Publication No. 2018/0285772 ) (“Gopalan”) and Karim et al. (U.S. Patent No. 11,755,951) (“Karim”).
Regarding claim 21, Tasinga in view of Gubin in further view of Gopalan teaches the system of claim 1, wherein the intelligent scheduling machine learned model is seeded with the initial retraining frequency (Gubin [0072] “The machine learning module 345 periodically retrains 630 the model corresponding to each subset of users of the social networking system. The machine learning module 345 may retrain each model at a fixed frequency.” Gubin teaches an initial/fixed/periodic retraining of a machine learning model for machine learning module 345, corresponding to seeding the model with an initial retraining frequency.) …and the intelligent scheduling machine learned model dynamically tunes a retraining frequency for the first inference model from the initial retraining frequency based on the set of features corresponding to the first entity (Gubin [0072] “The model corresponding to each subset may be retrained 630 at a different frequency compared to other models. The machine learning module 345 may use information describing how users of each subset interact with the news feed stories ranked by a model are presented to them to determine a rate at which the corresponding model is retrained 630.” Gubin teaches subsequently changing retraining frequency based on information describing how users of each subset interact with the news feed stories, corresponding to dynamically tuning the retraining frequency based on the set of features corresponding to an entity (i.e., information describing how users interact with the news feed stories).), but fails to explicitly teach …the initial retraining frequency being specified by the first entity.
However, Karim teaches the initial retraining frequency being specified by the first entity (Karim Col 4 lines 9-17 “The GUI 110 can allow the user to set configuration criteria 115 for continuous machine learning with respect to a model in a production environment. The user can access a screen for setting the criteria 115, in an example. The criteria 115 can include, for example, scheduling intervals for retraining a selected model using new datasets at a network location, such as daily, weekly, or monthly. The GUI 110 can also allow the user to specify a threshold accuracy level for a new model to be automatically deployed.” Karim teaches a use GUI to allow user to set a model retraining frequency, corresponding to an initial retraining frequency specified by an entity (user).),
Tasinga, Gubin, Gopalan and Karim are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically machine learning model retraining. Therefore, it would have obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tasinga in view of Gubin and Gopalan with the above teachings of Karim. Doing so can ensure that the model continuously updates with minimal user involvement, decreasing complexity, minimizing processing expense, all while maintaining the model's performance levels (Karim Col 2 lines 66-67, Col 3 lines 1-2 “This can ensure that the model continuously updates with minimal user involvement, decreasing complexity, minimizing processing expense, all while maintaining the model's performance levels.”).
Claim 22 is rejected under 35 U.S.C. 103 as being unpatentable over Tasinga et al. (U.S. Patent Publication No. 2022/0180178) (“Tasinga”) in view of Gubin et al. (U.S. Patent Publication No. 2013/0031489) (“Gubin”) in further view of Gopalan (U.S. Patent Publication No. 2018/0285772 ) (“Gopalan”) and Srinivasan et al. (U.S. Patent Publication No. 2018/0300653) (“Srinivasan”) in further view of Mathur (U.S. Patent Publication No. 2021/0281662) (“Mathur”).
Regarding claim 22, Tasinga in view of Gubin in further view of Gopalan and Srinivasan teaches the system of claim 7, as discussed above in the rejection of claim 7, but fails to explicitly teach wherein the dynamically assigning training of the first inference model to the first container is performed using a dynamic weighted container assignment machine learned model trained, using a clustering algorithm, to output a container configuration for a combination of an entity and an inference model.
However, Mathur teaches wherein the dynamically assigning training of the first inference model to the first container is performed using a dynamic weighted container assignment machine learned model trained, using a clustering algorithm, to output a container configuration for a combination of an entity and an inference model (Mathur [0026] “The deployment cluster stores this received model-specific configuration parameters in a local filesystem for access when needed. After the target model is identified, the deployment cluster can retrieve the model-specific configuration parameters from its local data storage (i.e., on the container file system associated with the host system of the deployment cluster). In various embodiments, the API server of the deployment cluster may retrieve the model-specific configuration parameters.”; [0031] “Using the configuration parameters received, the management system can then generate a serialized version of the machine learning model and store that version in the model container repository at operation 308.” Mathur teaches a clustering algorithm to retrieve the model-specific configuration parameters from a container file system, which is used to configure a container for a machine learning inference model.).
Tasinga, Gubin, Gopalan, Srinivasan and Mathur are all considered to be analogous to the claimed invention because they are in the same field of artificial intelligence and more specifically machine learning model training. Therefore, it would have obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Tasinga in view of Gubin, Gopalan and Srinivasan with the above teachings of Mathur. Doing so may reduce the downtime between adding or removing models from a cluster because the deployment engine can continue to run without the need to re-instantiate the API endpoints of the API server to tie it closely to a given model (Mathur [0013] “This reduces the downtime between adding or removing models from a cluster because the deployment engine can continue to run without the need to re-instantiate the API endpoints of the API server to tie it closely to a given model.”).
Response to Arguments
Regarding the rejection of the claims under 35 U.S.C. 101, the rejection of the Claims under 35 U.S.C. 101 has been reversed, as discussed in the “Patent Board Decision”, dated July 7, 2026.
Regarding the rejection of the claims under 35 U.S.C. 103, Applicant’s arguments with respect to the claims have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KURT NICHOLAS PRESSLY whose telephone number is (703)756-4639. The examiner can normally be reached M-F 8-4.
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, Kamran Afshar can be reached at (571) 272-7796. 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.
/KURT NICHOLAS PRESSLY/Examiner, Art Unit 2125
/KAMRAN AFSHAR/Supervisory Patent Examiner, Art Unit 2125
Prosecution Insights