DETAILED ACTION
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
2. A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. 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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office Action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 06/15/2026 has been entered.
3. Applicant's amendment and response received 06/15/2026, responding to the 11/18/2025 Office Action provided in the rejections of claims 1-20, wherein at least independent claims 1, 11 and 20 have been amended. Claims 1-20 remain pending in the application; which has been fully considered by the Examiner.
Response to Arguments
4. Applicant’s arguments with respect to newly amended independent claims 1, 11 and 20 and claims 2-10 and 12-19 on pages 7-13 of the response have been fully considered but they are not persuasive are moot in view of the new ground(s) of rejection- see Brown (Art newly made of record), Gold (Art of record) and Sidhu (Art newly made of record) as applied below, as they further teach such use.
Claim Rejections - 35 USC § 112
5. The following is a quotation of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
6. Claims 3, 12 and 13 are rejected under 35 U.S.C. 112(a) as failing to comply with the written description requirement. The claims contain subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor had possession of the claimed invention at the time the application was filed.
Claims 3 and 12 recites “the one or more processors implement a same GPU instruction set used to execute operations” (emphasis added). Applicant’s disclosure does not set forth any written description and is devoid of such “same GPU instruction set used” (emphasis added).
Claim 13 is also rejected for being dependent on rejected base claims.
Claim Rejections - 35 USC § 103
7. 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:
3A 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.
8. Claims 1, 3, 4, 11, 12 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Brown US 2018/0293463 in view of Gold et al., US 2019/0121889 (hereinafter Gold) in view of Sidhu et al., U.S. 2020/0034710 (hereinafter Sidhu).
In regards to claim 1, Brown teaches:
A method comprising determining that a version of one or more neural networks executed using one or more processors of a processor type exceeds an improvement benchmark compared to a different version of the one or more neural networks executed using the one or more processors of the same processor type (Abstract, see the learner module is configured to train a plurality of AI models in parallel, and the instructor module is configured to coordinate with a plurality of simulators for respectively training the AI models... The learner module is further configured to first train the AI models on a first batch of similar data synchronously pooled in a memory of the learner module with a first CPU, GPU, or DSP of one or more processors. The learner module is further configured to subsequently train the AI models on a second, different batch of similar data synchronously pooled in the memory of the learner module with the first CPU, GPU, or DSP) and (p. 13, [0128], see identify a best trained AI model (e.g., by means of optimal results based on factors such as performance time, accuracy, etc.) among the number of trained AI models) and (p. 14, [0135], see the hyperlearner module 225 can reference archived, previously built and trained intelligence models to help guide the instructor module 224 to train the current model of nodes. The hyperlearner module 225 can parse an archive database of trained intelligence models, known past similar problems and proposed solutions, and other sources. The hyperlearner module 225 can compare previous solutions similar to the solutions needed in a current problem as well as compare previous problems similar to the current problem to suggest potential optimal neural network topologies and training lessons and training methodologies) (emphasis added).
Brown doesn't explicitly teach:
deploying the
However, Gold teaches such use: (p. 18, [0149], see exploring parameters and models, quickly testing with a smaller dataset, and iterating to converge on the most promising models to push into the production cluster), (p. 51, [0371], see the example method depicted in FIG. 29 also includes identifying (2902), from amongst a plurality of machine learning models, a preferred machine learning model. Identifying (2902) a preferred machine learning model from amongst a plurality of machine learning models may be carried out, for example, by comparing a plurality of machine learning models to identify which machine learning model performed the best relative to a predetermined set of criteria (e.g., most accurate, quickest time to achieve a particular accuracy threshold, and so on. In such an example, a plurality of metrics may be used in a weighted or unweighted fashion to identify (2902) the preferred machine learning model. In such an example, the artificial intelligence infrastructure (2402) may be configured to use such information, for example, by automatically pushing the preferred machine learning model into a production environment), (p. 11, [0107], see storage clusters are connected to clients using Ethernet or fiber channel in some embodiments. If multiple storage clusters are configured into a storage grid, the multiple storage clusters are connected using the Internet) and (p. 14, [0128], see cloud services provider 302 may be configured to provide a variety of services to the storage system 306 and users of the storage system 306 … the cloud services provider 302 may be configured to provide services to the storage system 306 and users of the storage system 306 through the implementation of an infrastructure as a service (‘IaaS’) service model where the cloud services provider 302 offers computing infrastructure such as virtual machines and other resources as a service to subscribers” (emphasis added).
Brown and Gold are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown and Gold before him or her, to modify the system of Brown to include the teachings of Gold, as a system for ensuring reproducibility in an artificial intelligence infrastructure, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to collecting the labeled data that is crucial for training an accurate, as suggested by Gold (p. 51, [0371], p. 53, [0387]).
Brown and Gold, in particular Brown doesn't explicitly teach:
based, at least in part on a determination that inferences of the one or more neural networks for the one or more web-based service clients are executed using one or more processors of the same processor type.
However, Sidhu teaches such use: (Abstract, see the selected models are then trained and the performance of the trained models are analyzed. Based on the analysis of the performance of the trained models, a single model is selected for deployment to the target platform), (p. 5, [0052], see as such, a model that operates with a specific performance in a first platform having a first computer configuration, may not operate with acceptable performance in a second platform having a second computer configuration. As such, to deploy models across various platforms, different models are generated that are tailored to the capabilities of the respective platforms) and (p. 7, [0077], see the kernels are specifically designed for a platform. In some embodiments, kernels are implemented in the machine language or assembly language of a processor of the target platform) (emphasis added).
Brown, Gold and Sidhu are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold and Sidhu before him or her, to modify the system of Brown and Gold, in particular Brown to include the teachings of Sidhu, as a system for optimizing neural network structures, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to explore multiple different architectures, or multiple different configurations of the same architecture, as suggested by Sidhu (p. 5, [0052], p. 10, [0107]).
In regards to claim 3, Brown and Gold, in particular Brown doesn’t explicitly teach:
the determination that the one or more processors are of the same processor type comprises determining that the one or more processors implement a same GPU instruction set used to execute operations of the one or more neural networks.
However, Sidhu teaches such use: (Abstract, see the selected models are then trained and the performance of the trained models are analyzed. Based on the analysis of the performance of the trained models, a single model is selected for deployment to the target platform), (p. 5, [0052], see as such, a model that operates with a specific performance in a first platform having a first computer configuration, may not operate with acceptable performance in a second platform having a second computer configuration. As such, to deploy models across various platforms, different models are generated that are tailored to the capabilities of the respective platforms) and (p. 7, [0077], see the kernels are specifically designed for a platform. In some embodiments, kernels are implemented in the machine language or assembly language of a processor of the target platform) (emphasis added).
Brown, Gold and Sidhu are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold and Sidhu before him or her, to modify the system of Brown and Gold, in particular Brown to include the teachings of Sidhu, as a system for optimizing neural network structures, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to explore multiple different architectures, or multiple different configurations of the same architecture, as suggested by Sidhu (p. 5, [0052], p. 10, [0107]).
In regards to claim 4, Brown doesn't explicitly teach:
the deployed version of the one or more neural networks meet higher improvement benchmark thresholds than a version currently used by the one or more web-based service users.
However, Gold teaches such use: (p. 51, [0371], see FIG. 29 also includes identifying (2902), from amongst a plurality of machine learning models, a preferred machine learning model. Identifying (2902) a preferred machine learning model from amongst a plurality of machine learning models may be carried out, for example, by comparing a plurality of machine learning models to identify which machine learning model performed the best relative to a predetermined set of criteria (e.g., most accurate, quickest time to achieve a particular accuracy threshold, and so on. In such an example, a plurality of metrics may be used in a weighted or unweighted fashion to identify (2902) the preferred machine learning model. In such an example, the artificial intelligence infrastructure (2402) may be configured to use such information, for example, by automatically pushing the preferred machine learning model into a production environment… providing a recommendation that a particular user should switch from one machine learning model to another machine learning model in their production environment, by automatically rolling back to a previous version of a machine learning model if it is determined to be preferred over a subsequent version of the machine learning model, and so on. In the example method depicted in FIG. 29, identifying (2902) a preferred machine learning model from amongst a plurality of machine learning models can include evaluating (2904) one or more machine learning models utilizing a predetermined model test. The predetermined model test may be embodied, for example, as an A/B testing model in which two machine learning models are compared) (emphasis added).
Brown and Gold are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown and Gold before him or her, to modify the system of Brown to include the teachings of Gold, as a system for ensuring reproducibility in an artificial intelligence infrastructure, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to collecting the labeled data that is crucial for training an accurate, as suggested by Gold (p. 51, [0371], p. 53, [0387]).
In regards to claim 11, Brown teaches:
A non-transitory computer-readable medium storing computer instructions that, when executed by one or more processors, cause the one or more processors to perform a method comprising: determining that a version of one or more neural networks executed using one or more processors of a processor type exceeds an improvement benchmark compared to a different version of the one or more neural networks executed using the one or more processors of the same processor type (Abstract, see the learner module is configured to train a plurality of AI models in parallel, and the instructor module is configured to coordinate with a plurality of simulators for respectively training the AI models... The learner module is further configured to first train the AI models on a first batch of similar data synchronously pooled in a memory of the learner module with a first CPU, GPU, or DSP of one or more processors. The learner module is further configured to subsequently train the AI models on a second, different batch of similar data synchronously pooled in the memory of the learner module with the first CPU, GPU, or DSP) and (p. 13, [0128], see identify a best trained AI model (e.g., by means of optimal results based on factors such as performance time, accuracy, etc.) among the number of trained AI models) and (p. 14, [0135], see the hyperlearner module 225 can reference archived, previously built and trained intelligence models to help guide the instructor module 224 to train the current model of nodes. The hyperlearner module 225 can parse an archive database of trained intelligence models, known past similar problems and proposed solutions, and other sources. The hyperlearner module 225 can compare previous solutions similar to the solutions needed in a current problem as well as compare previous problems similar to the current problem to suggest potential optimal neural network topologies and training lessons and training methodologies) (emphasis added).
Brown doesn't explicitly teach:
deploying the
However, Gold teaches such use: (p. 18, [0149], see exploring parameters and models, quickly testing with a smaller dataset, and iterating to converge on the most promising models to push into the production cluster), (p. 51, [0371], see the example method depicted in FIG. 29 also includes identifying (2902), from amongst a plurality of machine learning models, a preferred machine learning model. Identifying (2902) a preferred machine learning model from amongst a plurality of machine learning models may be carried out, for example, by comparing a plurality of machine learning models to identify which machine learning model performed the best relative to a predetermined set of criteria (e.g., most accurate, quickest time to achieve a particular accuracy threshold, and so on. In such an example, a plurality of metrics may be used in a weighted or unweighted fashion to identify (2902) the preferred machine learning model. In such an example, the artificial intelligence infrastructure (2402) may be configured to use such information, for example, by automatically pushing the preferred machine learning model into a production environment), (p. 11, [0107], see storage clusters are connected to clients using Ethernet or fiber channel in some embodiments. If multiple storage clusters are configured into a storage grid, the multiple storage clusters are connected using the Internet) and (p. 14, [0128], see cloud services provider 302 may be configured to provide a variety of services to the storage system 306 and users of the storage system 306 … the cloud services provider 302 may be configured to provide services to the storage system 306 and users of the storage system 306 through the implementation of an infrastructure as a service (‘IaaS’) service model where the cloud services provider 302 offers computing infrastructure such as virtual machines and other resources as a service to subscribers” (emphasis added).
Brown and Gold are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown and Gold before him or her, to modify the system of Brown to include the teachings of Gold, as a system for ensuring reproducibility in an artificial intelligence infrastructure, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to collecting the labeled data that is crucial for training an accurate, as suggested by Gold (p. 51, [0371], p. 53, [0387]).
Brown and Gold, in particular Brown doesn't explicitly teach:
based, at least in part on a determination that inferences of the one or more neural networks for the one or more web-based service clients are executed using one or more processors of the same processor type.
However, Sidhu teaches such use: (Abstract, see the selected models are then trained and the performance of the trained models are analyzed. Based on the analysis of the performance of the trained models, a single model is selected for deployment to the target platform), (p. 5, [0052], see as such, a model that operates with a specific performance in a first platform having a first computer configuration, may not operate with acceptable performance in a second platform having a second computer configuration. As such, to deploy models across various platforms, different models are generated that are tailored to the capabilities of the respective platforms) and (p. 7, [0077], see the kernels are specifically designed for a platform. In some embodiments, kernels are implemented in the machine language or assembly language of a processor of the target platform) (emphasis added).
Brown, Gold and Sidhu are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold and Sidhu before him or her, to modify the system of Brown and Gold, in particular Brown to include the teachings of Sidhu, as a system for optimizing neural network structures, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to explore multiple different architectures, or multiple different configurations of the same architecture, as suggested by Sidhu (p. 5, [0052], p. 10, [0107]).
In regards to claim 12, Brown and Gold, in particular Brown doesn’t explicitly teach:
the determination that the one or more processors are of the same processor type comprises determining that the one or more processors implement a same GPU instruction set used to execute operations of the one or more neural networks.
However, Sidhu teaches such use: (Abstract, see the selected models are then trained and the performance of the trained models are analyzed. Based on the analysis of the performance of the trained models, a single model is selected for deployment to the target platform), (p. 5, [0052], see as such, a model that operates with a specific performance in a first platform having a first computer configuration, may not operate with acceptable performance in a second platform having a second computer configuration. As such, to deploy models across various platforms, different models are generated that are tailored to the capabilities of the respective platforms) and (p. 7, [0077], see the kernels are specifically designed for a platform. In some embodiments, kernels are implemented in the machine language or assembly language of a processor of the target platform) (emphasis added).
Brown, Gold and Sidhu are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold and Sidhu before him or her, to modify the system of Brown and Gold, in particular Brown to include the teachings of Sidhu, as a system for optimizing neural network structures, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to explore multiple different architectures, or multiple different configurations of the same architecture, as suggested by Sidhu (p. 5, [0052], p. 10, [0107]).
In regards to claim 13, Brown teaches:
The processor type corresponds to a processor of a graphics processing unit (Abstract, see the learner module is configured to train a plurality of AI models in parallel, and the instructor module is configured to coordinate with a plurality of simulators for respectively training the AI models... The learner module is further configured to first train the AI models on a first batch of similar data synchronously pooled in a memory of the learner module with a first CPU, GPU, or DSP of one or more processors. The learner module is further configured to subsequently train the AI models on a second, different batch of similar data synchronously pooled in the memory of the learner module with the first CPU, GPU, or DSP) (emphasis added).
In regards to claim 20, Brown teaches:
A system, comprising: a memory storing instructions; and one or more processors that execute the instructions to perform a method comprising: determining that a version of one or more neural networks executed using one or more processors of a processor type exceeds an improvement benchmark compared to a different version of the one or more neural networks executed using the one or more processors of the same processor type (Abstract, see the learner module is configured to train a plurality of AI models in parallel, and the instructor module is configured to coordinate with a plurality of simulators for respectively training the AI models... The learner module is further configured to first train the AI models on a first batch of similar data synchronously pooled in a memory of the learner module with a first CPU, GPU, or DSP of one or more processors. The learner module is further configured to subsequently train the AI models on a second, different batch of similar data synchronously pooled in the memory of the learner module with the first CPU, GPU, or DSP) and (p. 13, [0128], see identify a best trained AI model (e.g., by means of optimal results based on factors such as performance time, accuracy, etc.) among the number of trained AI models) and (p. 14, [0135], see the hyperlearner module 225 can reference archived, previously built and trained intelligence models to help guide the instructor module 224 to train the current model of nodes. The hyperlearner module 225 can parse an archive database of trained intelligence models, known past similar problems and proposed solutions, and other sources. The hyperlearner module 225 can compare previous solutions similar to the solutions needed in a current problem as well as compare previous problems similar to the current problem to suggest potential optimal neural network topologies and training lessons and training methodologies) (emphasis added).
Brown doesn't explicitly teach:
deploying the
However, Gold teaches such use: (p. 18, [0149], see exploring parameters and models, quickly testing with a smaller dataset, and iterating to converge on the most promising models to push into the production cluster), (p. 51, [0371], see the example method depicted in FIG. 29 also includes identifying (2902), from amongst a plurality of machine learning models, a preferred machine learning model. Identifying (2902) a preferred machine learning model from amongst a plurality of machine learning models may be carried out, for example, by comparing a plurality of machine learning models to identify which machine learning model performed the best relative to a predetermined set of criteria (e.g., most accurate, quickest time to achieve a particular accuracy threshold, and so on. In such an example, a plurality of metrics may be used in a weighted or unweighted fashion to identify (2902) the preferred machine learning model. In such an example, the artificial intelligence infrastructure (2402) may be configured to use such information, for example, by automatically pushing the preferred machine learning model into a production environment), (p. 11, [0107], see storage clusters are connected to clients using Ethernet or fiber channel in some embodiments. If multiple storage clusters are configured into a storage grid, the multiple storage clusters are connected using the Internet) and (p. 14, [0128], see cloud services provider 302 may be configured to provide a variety of services to the storage system 306 and users of the storage system 306 … the cloud services provider 302 may be configured to provide services to the storage system 306 and users of the storage system 306 through the implementation of an infrastructure as a service (‘IaaS’) service model where the cloud services provider 302 offers computing infrastructure such as virtual machines and other resources as a service to subscribers” (emphasis added).
Brown and Gold are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown and Gold before him or her, to modify the system of Brown to include the teachings of Gold, as a system for ensuring reproducibility in an artificial intelligence infrastructure, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to collecting the labeled data that is crucial for training an accurate, as suggested by Gold (p. 51, [0371], p. 53, [0387]).
Brown and Gold, in particular Brown doesn't explicitly teach:
based, at least in part on a determination that inferences of the one or more neural networks for the one or more web-based service clients are executed using one or more processors of the same processor type.
However, Sidhu teaches such use: (Abstract, see the selected models are then trained and the performance of the trained models are analyzed. Based on the analysis of the performance of the trained models, a single model is selected for deployment to the target platform), (p. 5, [0052], see as such, a model that operates with a specific performance in a first platform having a first computer configuration, may not operate with acceptable performance in a second platform having a second computer configuration. As such, to deploy models across various platforms, different models are generated that are tailored to the capabilities of the respective platforms) and (p. 7, [0077], see the kernels are specifically designed for a platform. In some embodiments, kernels are implemented in the machine language or assembly language of a processor of the target platform) (emphasis added).
Brown, Gold and Sidhu are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold and Sidhu before him or her, to modify the system of Brown and Gold, in particular Brown to include the teachings of Sidhu, as a system for optimizing neural network structures, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to explore multiple different architectures, or multiple different configurations of the same architecture, as suggested by Sidhu (p. 5, [0052], p. 10, [0107]).
9. Claims 2, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Brown in view of Gold in view of Sidhu in view of Brueckner et al., US 2016/0078361 (hereinafter Brueckner).
In regards to claims 1 and 11, the rejections above are incorporated respectively.
In regards to claim 2, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
the one or more neural networks are to be used to perform inferencing operations and to output inferenced data to a software application installed on a client computing device.
However, Brueckner teaches such use: (Fig. 1, see process arrow flow from Raw data 130, Input record handlers 160, Feature Processors 162, ML Algorithm implementations 166, Workload distribution 175, server pools 185, MlS artifact repository 120, Clients 164, Machine learning service 189) and (p. 6, [0064}, see the output 116 of the feature processing transformations may in turn be used as input for a selected machine learning algorithm 166, which may be executed in accordance with algorithm parameters 154 using yet another set of resources from pool 185. A wide variety of machine learning algorithms may be supported natively by the MLS libraries, including for example random forest algorithms, neural network algorithms, stochastic gradient descent algorithms, and the like).
Brown, Gold, Sidhu and Brueckner are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Brueckner before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Brueckner, as a system for training machine learning models, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data transformations to apply to data input, as suggested by Brueckner (p. 6, [0064}, p 50, [0562-0564]).
In regards to claim 17, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
the one or more neural networks perform inferencing operations and provide inferenced data to a software application by at least: providing input data by the software application to the one or more neural networks, processing the input data by the one or more neural networks to generate the inferenced data, and outputting the inferenced data to the software application.
However, Brueckner teaches such use: (Fig. 1, see process arrow flow from Raw data 130, Input record handlers 160, Feature Processors 162, ML Algorithm implementations 166, Workload distribution 175, server pools 185, MlS artifact repository 120, Clients 164, Machine learning service 189) and (p. 6, [0064}, see the output 116 of the feature processing transformations may in turn be used as input for a selected machine learning algorithm 166, which may be executed in accordance with algorithm parameters 154 using yet another set of resources from pool 185. A wide variety of machine learning algorithms may be supported natively by the MLS libraries, including for example random forest algorithms, neural network algorithms, stochastic gradient descent algorithms, and the like).
Brown, Gold, Sidhu and Brueckner are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Brueckner before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Brueckner, as a system for training machine learning models, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data transformations to apply to data input, as suggested by Brueckner (p. 6, [0064}, p 50, [0562-0564]).
10. Claims 5-10, 15-16 and 18 are rejected under as being unpatentable over Brown in view of Gold in view of Sidhu in view of Novielli et al., U.S. 2019/0278870 (hereinafter Novielli).
In regards to claims 1 and 11 the rejections above are incorporated respectively.
In regards to claim 5, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
further comprising generating the version of the one or more neural networks by at least retraining the one or more neural networks.
However, Novielli teaches such use: (p. 3, [0050], see due to the number of queries processed each second by the search system 104 and the amount of feedback received by the machine learning model creation system 106, a sufficient amount of data to retrain the model can be received very quickly. Thus, the machine learning model can be retrained, a new model created… In one embodiment, the trigger event can be time so that the machine learning model is updated (i.e., an existing model retrained or a new model created) on a periodic or a periodic schedule. In another embodiment, the trigger even can be machine learning model performance so that the machine learning model is updated when the performance falls below a threshold).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Gold to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
In regards to claim 6, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
automatically distributing the one or more updates to the one or more neural networks by at least distributing one or more updated parameters the one or more neural networks to a client computing device.
However, Novielli teaches such use: (p. 3, [0050] in yet another embodiment, performance of a test machine learning model triggers updating the machine learning model to a wider group of users (see the A/B testing discussion below). Thus, when a test machine learning model (i.e., model A) performance exceeds another deployed model (i.e., model B), the A model can be deployed to replace the B model. In other embodiments a combination of trigger events can be used) and (p. 4, [0053], see the model creation process 122 can distribute a whole new model, updated coefficients for an existing model, or both… Coefficients in this context are the weights or other parameters that turn an untrained learning model into a trained learning model) and (p. 4, [0055], see various machine learning models that receive an input list and output an output list can be used in embodiments of the current disclosure … appropriate models are those that are able to classify and/or rank a list of input items, including supervised, unsupervised and semi-supervised models. Various models fall into the categories such as deep learning models, ensemble models, neural networks) (emphasis added).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
In regards to claim 7, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
the automatically distributing the one or more updates to the one or more neural networks comprises distributing one or more updated hyperparameter adjustments for the one or more neural networks to the client computing device.
However, Novielli teaches such use: (p. 3, [0050] in yet another embodiment, performance of a test machine learning model triggers updating the machine learning model to a wider group of users (see the A/B testing discussion below). Thus, when a test machine learning model (i.e., model A) performance exceeds another deployed model (i.e., model B), the A model can be deployed to replace the B model. In other embodiments a combination of trigger events can be used) and (p. 4, [0053], see the model creation process 122 can distribute a whole new model, updated coefficients for an existing model, or both… Coefficients in this context are the weights or other parameters that turn an untrained learning model into a trained learning model).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
In regards to claim 8, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
the automatically distributing the one or more updates to the one or more neural networks comprises updating the one or more neural networks of the client computing device with at least one of a layer substitution, a layer fusing, or input stacking.
However, Novielli teaches such use: (p. 3, [0050] in yet another embodiment, performance of a test machine learning model triggers updating the machine learning model to a wider group of users (see the A/B testing discussion below). Thus, when a test machine learning model (i.e., model A) performance exceeds another deployed model (i.e., model B), the A model can be deployed to replace the B model. In other embodiments a combination of trigger events can be used).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
In regards to claim 9, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
generating the version of the one or more neural networks wherein automatically.
However, Novielli teaches such use: (p. 3, [0050] in yet another embodiment, performance of a test machine learning model triggers updating the machine learning model to a wider group of users (see the A/B testing discussion below). Thus, when a test machine learning model (i.e., model A) performance exceeds another deployed model (i.e., model B), the A model can be deployed to replace the B model. In other embodiments a combination of trigger events can be used) and (p. 4, [0055], see various machine learning models that receive an input list and output an output list can be used in embodiments of the current disclosure… appropriate models are those that are able to classify and/or rank a list of input items, including supervised, unsupervised and semi-supervised models. Various models fall into the categories such as deep learning models, ensemble models, neural networks) (emphasis added).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
In regards to claim 10, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
automatically distributing the version the one or more neural networks to the one or more web-based service client when an updated version of the one or more neural networks meets or exceeds one or more thresholds for improvement relating to at least one of accuracy, quality, or performance.
However, Novielli teaches such use: (p. 3, [0050] In another embodiment, the trigger even can be machine learning model performance so that the machine learning model is updated when the performance falls below a threshold such as when the number or rate of incorrect prefetches exceeds a threshold and/or the number or rate of correct prefetches falls below a threshold. In yet another embodiment, performance of a test machine learning model triggers updating the machine learning model to a wider group of users (see the A/B testing discussion below). Thus, when a test machine learning model (i.e., model A) performance exceeds another deployed model (i.e., model B), the A model can be deployed to replace the B model. In other embodiments a combination of trigger events can be used).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
In regards to claim 15, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
a software application is a voice recognition application that uses the one or more neural networks.
However, Novielli teaches such use: (p. 4, [0063], see input is received in operation 306. As noted above, the input can be in any format such as text, voice, gesture, and so forth).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
In regards to claim 16, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
the one or more neural networks generates inferenced data that includes one or more audio-related inferences, the one or more audio-related inferences being at least one of a language translation or a voice recognized command.
However, Novielli teaches such use: (p. 4, [0063], see input is received in operation 306. As noted above, the input can be in any format such as text, voice, gesture, and so forth).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
In regards to claim 18, Brown, Gold and Sidhu, in particular Brown doesn’t explicitly teach:
the version of the one or more neural networks is received responsive to a trigger.
However, Novielli teaches such use: (p. 3, [0050] in yet another embodiment, performance of a test machine learning model triggers updating the machine learning model to a wider group of users (see the A/B testing discussion below). Thus, when a test machine learning model (i.e., model A) performance exceeds another deployed model (i.e., model B), the A model can be deployed to replace the B model. In other embodiments a combination of trigger events can be used) and (p. 4, [0055], see various machine learning models that receive an input list and output an output list can be used in embodiments of the current disclosure… appropriate models are those that are able to classify and/or rank a list of input items, including supervised, unsupervised and semi-supervised models. Various models fall into the categories such as deep learning models, ensemble models, neural networks) (emphasis added).
Brown, Gold, Sidhu and Novielli are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Novielli before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Novielli, as a system for loading search results using machine learning, and accordingly it would enhance the system Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize data collected from user devices as feedback to train machine learning models, as suggested by Novielli (p. 3, [0050], p. 11, [0209]).
11. Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over Brown in view of Gold in view of Sidhu in view of Brueckner in view of Skuin et al., U.S. Patent No. 10,7135,43 (hereinafter Skuin).
In regards to claim 11 the rejections above are incorporated respectively.
In regards to claim 14, Brown, Gold, Sidhu and Brueckner, in particular Brown doesn’t explicitly teach:
the one or more neural networks are used to generate inferenced data that includes one or more image-related inferences, the one or more image-related inferences being at least one of an anti-aliased image, an images with upscaled resolution, or a denoised image.
However, Skuin teaches such use: (column 11, lines 12-21, see to learn detail of these video game characters it may be beneficial for the machine learning models to have access to high resolution and/or large versions of the characters. Thus, the training data 112 may include the particular video game characters rendered at different scales. Additionally, the different scales may be beneficial as real-world gameplay (e.g., a broadcast of a hockey game) may include images captured by television cameras at different zoom levels).
Brown, Gold, Sidhu, Brueckner and Skuin are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu, Brueckner and Skuin before him or her, to modify the system of Brown, Gold, Sidhu and Brueckner, in particular Brown to include the teachings of Skuin, as a system for machine training, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to utilize machine learning with a video game, as suggested by Skuin (column 11, lines 12-21, p. 29, lines 41-46).
12. Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over Brown in view of Gold in view of Sidhu view of Palladino et al., U.S. 2019/0037005 (hereinafter Palladino).
In regards to claim 11 the rejections above are incorporated respectively.
In regards to claim 19, Brown, Gold, Sidhu and Brueckner, in particular Brown doesn’t explicitly teach:
request is triggered as a result of a call to a feature application programming interface (API) that causes execution of the one or more neural networks.
However, Palladino teaches such use: (p. 10, 2nd column, 1st para., see the system of claim 17, wherein the executed auto-documentation code is further configured to update the machine learning model based on the content of the API request or response as observed).
Brown, Gold, Sidhu and Palladino are analogous art because they are from the same field of endeavor, AI model updates.
Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teaching of Brown, Gold, Sidhu and Palladino before him or her, to modify the system of Brown, Gold and Sidhu, in particular Brown to include the teachings of Palladino, as a system for processing machine learning API request, and accordingly it would enhance the system of Brown, which is focused on artificial intelligence with enhanced hardware, because that would provide Brown with the ability to update a machine learning model, as suggested by Palladino (p. 10, 2nd column, p. 9, [0093]).
Conclusion
13. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US Patent Application Publications
Abeysooriya 9336483 teaches updating neural network systems may be implemented to generate, train, evaluate and update artificial neural network data structures used by content distribution networks. Such systems and methods described herein may include generating and training neural networks, using neural networks to perform predictive analysis and other decision-making processes within content distribution networks, evaluating the performance of neural networks, and generating and training pluralities of replacement candidate neural networks within cloud computing architectures and/or other computing environments.
Li 12393835 teaches a performance benchmark database is obtained, where the performance benchmark database at least includes structural data of one or more deep neural network models, time performance data and computing resource consumption data of a plurality of deep learning applications based on the one or more deep neural network models; a training dataset is extracted based on the performance benchmark database, where the training dataset has a plurality of parameter dimensions, the plurality of parameter dimensions including: structures of deep neural network models of the plurality of deep learning applications, resource configuration of the plurality of deep learning applications, and training time of the plurality of deep learning applications; and correspondence among the parameter dimensions of the training dataset is created so as to create an estimation model for estimating resources utilized by deep learning applications.
14. Any inquiry concerning this communication or earlier communications from the examiner should be directed to Evral Bodden whose telephone number is 571-272-3455. The examiner can normally be reached on Monday to Friday from 9am to 5pm.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Chat Do, can be reached at telephone number 571-272-3721. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
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) Form at https://www.uspto.gov/patents/uspto-automatedinterview-request-air-form.
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.
/EVRAL E BODDEN/Primary Examiner, Art Unit 2193
Prosecution Insights