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 . 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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 12/17/2025 has been entered.
Remarks
This Office Action is responsive to Applicants' Amendment filed on December 17, 2025, in which claims 1, 11, and 20 are currently amended. Claims 1-20 are currently pending.
Response to Arguments
The rejections to claims 1, 11, and 20 under 35 U.S.C. § 112(b) are hereby withdrawn, as necessitated by applicant's amendments and remarks made to the rejections.
Applicant’s arguments with respect to rejection of claims 1-20 under 35 U.S.C. 101 based on amendment have been considered and are persuasive. The rejections to claims 1-20 under 35 U.S.C. § 101 are hereby withdrawn, as necessitated by applicant's amendments and remarks made to the rejections.
Applicant’s arguments with respect to rejection of claims 1-20 under 35 U.S.C. 103 based on amendment have been considered and are persuasive. The argument is moot in view of a new ground of rejection set forth below.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claims 1, 11, and 20, "identifying at least one of a lack of generality, a lack of robustness” is indefinite. “a lack of generality” and “a lack of robustness” are terms of degrees with no defined constraints. One of ordinary skill in the art would recognize that in machine learning “generality”/generalization is an abstract, non-unitary concept with multiple conflicting measurement methods (generality can be, but is not required to be measured by one or more of: accuracy, precision/recall, F1 score, ROC AUC, mean absolute error, mean squared error, R-squared, learning curves, out-of-distribution training, and benchmark datasets. This list is not exhaustive and other methods exist for attempting to measure a model’s “generality”) and similarly “generality” may be defined on any of the above methods tested on new but IID data, distribution shift, cross-domain transfer, robustness to perturbations, task-level generalization in meta-learning, etc. A person of ordinary skill in the art would recognize that generality is a contested, context dependent concept with no “default” test that everyone would apply absent instructions. Since generality is not defined or limited in the instant specification at all and multiple contradictory boundaries and measurements for generality exist in the art, the scope of the claim cannot be reasonably determined. The same is also true for “robustness”.
Regarding claims 3 and 13, "an inference that the artificial intelligence model fails to meet required generality or robustness" is indefinite. It would not be clear to one of ordinary skill in the art specifically how to make a determination/inference that an artificial intelligence model fails to meet generality or robustness. The scope of the term "generality" is especially unclear such that the scope of the claim cannot be reasonably determined. In the interest of further examination the claim limitation is interpreted as "an inference about the performance of the artificial intelligence model".
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 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.
Claims 1-4, 7, 10-14, 17, and 20 are rejected under U.S.C. §103 as being unpatentable over the combination of Polleri (US20210081819A1) and Gardner (US20190286086A1).
PNG
media_image1.png
648
480
media_image1.png
Greyscale
FIG. 1 of US20210081819A1
Regarding claim 1, Polleri teaches A method comprising: obtaining, by a device comprising at least one processor ([¶0052] "The model execution engine 108 can execute the machine learning application 112 on infrastructure 128 using one or more the infrastructure interfaces 124. The infrastructure 128 can include one or more processors, one or more memories, and one or more network interfaces, one or more buses and control lines that can be used to generate, test, compile, and deploy a machine learning application 112" [¶0433] "FIG. 22 illustrates an exemplary computer system 2200, in which various embodiments of the present invention may be implemented. The system 2200 may be used to implement any of the computer systems described above. As shown in the figure, computer system 2200 includes a processing unit 2204 that communicates with a number of peripheral subsystems via a bus subsystem 2202")
configured to execute a neuro-symbolic metamodel ([¶0339] "A. Ontology Modeling and Building Mechanism 1608" [[¶0340] "The semantic profile for machine learning platform 100 can include: functional semantics of each microservices routine 140 [...]provenance of services that are composed as it can affect the performance at run time of the machine learning application 112)" [¶0341] "B. Reasoner Engine 1604 to Process the Ontology 1616" [¶0342] "The reasoner engine 1604 guides the search for the best combination of components (e.g., model, metrics) to solve a problem. The reasoner engine 1604 determines the software functions and hardware to combine into the product graph 1620." [¶0392] "Machine learning services and their ontologies are defined in deployable service descriptions, which are used by the model composition engine 132 to assemble a composite service to trigger search for the best architectural model for run-time. The architectural model includes a pipeline 136 specifying any microservices routines 140, software modules 144, and infrastructure modules 148 along with any customizations and interdependencies" Polleri explicitly links the Ontology Modeling (Reasoner engine interpreted as the neuro-symbolic metamodel) for execution using the ML Platform 100 through pipelines 136)
PNG
media_image2.png
648
474
media_image2.png
Greyscale
FIG. 16 of US20210081819A1
and to communicate with a training system,([¶0270] "a code integration request prediction server (or prediction server) 1010 may communicate with various client devices 1050, software development environments 1020, and other various systems over one or more communication networks 1040, to generate and train machine learning models as well as to use the trained models to predict code integration request outcomes [...] The prediction server 1010, discussed in more detail below, may include various hardware and/or software systems and sub-components, including trained machine-learning models 1015 as well as model training systems 1016 and model execution systems 1018" [¶0279] "it should be understood that the techniques described in reference to FIG. 10 need not be tied to any particular devices or servers within the computing environment 1000, but may be implemented by any computing systems and devices described or supported herein." Polleri explicitly discloses that FIG. 10 may be linked to other systems described such as FIG. 1 which shows a significant amount of overlap (see Model training system, data storage, and Model execution systems which directly map to FIG. 1). In other words FIG. 1 is the infrastructure and FIG. 10 is the training/execution applications whose functions can be realized on FIG. 1 or FIG. 22 which Polleri explicitly states may be used to execute any of the described computer systems)
data regarding generation of an artificial intelligence model by the training system, the data comprising training data used to train the artificial intelligence model, ([¶0105] "A model training system 1016 may retrieve data from data stores 1030 and/or client systems 1050, in order to train models 115 to generate predictive outcomes")
PNG
media_image3.png
524
706
media_image3.png
Greyscale
FIG. 10 of US20210081819A1
hyperparameters of the artificial intelligence model,([¶0377] "there is another kind of parameters that cannot be directly learned from the regular training process. These parameters express “higher-level” properties of the model such as its complexity or how fast it should learn. These are called hyperparameters. Hyperparameters are usually fixed before the actual training process begins. A hyperparameter is a configuration that is external to the model")
and an intended deployment environment for the artificial intelligence model;([¶0277] "Data repositories 1030 may include databases or data store structures storing various data relating to previous (or historical) code integration requests. Such historical data may include data detailing the particular characteristics of each code integration request (e.g. [...] the planned deployment environment" [¶0105] " The model composition engine 132 can receive several other user inputs including a second input identifying a data source for the machine learning architecture and a third input of one or more constraints (e.g., resources, location, security, or privacy) for the machine learning architecture")
analyzing, by the device, the data using a neuro-symbolic metamodel, to match the data to one or more concepts of a knowledge graph of the neuro-symbolic metamodel,([¶0319] "At 1510, the functionality can include accessing a library of terms stored in a memory, wherein the terms correspond to categories known by a machine learning model" [¶0326] "the method can include presenting the mapping of the one or more labels with the categories of the machine learning model. The technique can include receiving a second input. The second input can correlate to a label of the one or more labels for the schema of the data with a category of the one or more categories known by the machine learning model" [¶0342] "Using the created ontology 1616 (description on how the service internally works in terms of the interplay between data and control flow, and QoS benchmarks) for annotating services with concepts which are defined in formal logic-based ontologies such that, from a machine learning perspective, intelligent agents and a reasoner engine 1604 can determine formal service semantics and compose them based-on optimal run-time expectation by the product" [¶0346] "Product graphs 1620 are visual representations of data that represent mathematical structures used to study pairwise relationships between objects and entities. In computing, a graph database (GDB) is a database that uses graph structures for semantic queries with nodes, edges, and properties to represent and store data. A key concept of the system is the graph (or edge or relationship)" See also FIG. 16 which explicitly relates the Ontologies, Reasoner Engine, and Product Graphs)
making, by the device, one or more inferences about the artificial intelligence model, identifying at least one of a lack of generality, a lack of robustness, or a lack of accuracy of the artificial intelligence model,([¶0047] "A monitoring engine 156 can monitor operation of the machine learning applications 112 according to the KPI/QoS metrics 160 to assure the machine learning application 112 is performing according to requirements" [¶0073] "The training metrics help evaluate the performance of the model. Example training metrics can include classification accuracy, logarithmic loss, area under curve, F1 Score, mean absolute error, and mean squared error.")
by applying a semantic reasoning engine to the one or more concepts of the knowledge graph; and([¶0355] "the technique can learn from previous ontologies that certain KPIs and metrics are effective for solving certain solutions. Therefore the technique can recommend incorporating various KPIs or metrics into the product graph.")
causing, by the device and based on the one or more inferences, generation of a replacement artificial intelligence model by the training system ([¶0047] "A monitoring engine 156 can monitor operation of the machine learning applications 112 according to the KPI/QoS metrics 160 to assure the machine learning application 112 is performing according to requirements. In addition the monitoring engine 156 can seamlessly test end-to-end a new or evolving machine learning application at different scales, settings, loading, settings, etc")
to replace the artificial intelligence model by transmitting control commands via a network interface to the training system to use altered training data or altered hyperparameters identified by the neuro-symbolic metamodel, thereby improving computational efficiency or accuracy of the artificial intelligence model when deployed in the intended deployment environment.([¶0014] "The machine learning application with the new model can be tested off-line with the results being compared with ground truth data. If the machine learning application with the new model outperforms the previously used model, the machine learning application can be upgraded and auto-promoted to production. One or more parameters may also be discovered. The new parameters may be incorporated into the existing model in an off-line mode. The machine learning application with the new parameters can be tested off-line and the results can be compared with previous results with existing parameters. If the new parameters outperform the existing parameters as compared with ground-truth data, the machine learning application can be auto-promoted to production").
While Polleri teaches ([[¶0333] "An ontology consists of classes hierarchically arranged in a taxonomy of subclass-superclass, slots with descriptions defining value constraints, and values for these slots" [¶0346] "Product graphs 1620 are visual representations of data that represent mathematical structures used to study pairwise relationships between objects and entities. In computing, a graph database (GDB) is a database that uses graph structures for semantic queries with nodes, edges, and properties to represent and store data. A key concept of the system is the graph (or edge or relationship)" [¶0356] "At 1712, the functionality can include composing a product graph based on the on the first ontology, the one or more constraints, and one or more previous product graphs stored in the memory. The product graph relates the one or more data objects to a collection of nodes and edges for the data") which under a broad interpretation could reasonably be interpreted such that Subclass-superclass interpreted as synonymous with symbolic/sub-symbolic concept/data relationships. For example Polleri's explicit application of the system to eye images has a symbolic abstraction layer (ontology categories/concepts/taxonomy i.e. "eye images") and a sub-symbolic processing layer ("indicating the level of retinopathy"). See also FIG. 16. Examiner believers that Polleri does not explicitly teach the knowledge graph having multiple levels of abstraction ranging from sub-symbolic data to symbolic concepts.
Gardner, in the same field of endeavor, teaches the knowledge graph having multiple levels of abstraction ranging from sub-symbolic data to symbolic concepts ([¶0371] "constraints draw upon both symbolic and sub-symbolic aspects of the underlying knowledge graph, with both logical/linguistic relationships, and underlying contextual semantic distances between concepts being represented in the same graph" [¶098] "Such probabilistic semantic distance metrics are susceptible to being represented as relationships between two concepts in the semantically normalized knowledge graph and used to determine the degree of connectedness of concepts above").
Polleri's platform is about semantic, context-aware composition using ontology semantics and a reasoner engine. Gardner teaches that adding sub-symbolic semantic distances to a knowledge graph helps constrain/search more effectively because the KG includes both logical relationships and contextual semantic distances "in the same graph". Therefore, Polleri as well as Gardner are reasonably pertinent analogous art. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Polleri with the teachings of Gardner by using the KG having symbolic and sub-symbolic structure as the KG in Polleri. Polleri's platform is about semantic, context-aware composition using ontology semantics and a reasoner engine. Gardner provides as additional motivation for combination that the improved knowledge graph allows a user ([¶0371] “to use constraints and object functions to search for specifically defined properties”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 2, the combination of Polleri and Gardner teaches The method as in claim 1, wherein the replacement artificial intelligence model comprises a deep learning model.(Polleri [¶0282] "At 1106, the model training system 1016 (or other components with the prediction server 1010) may generate one or more model data structures, and at 1106 the models may be trained using machine-learning algorithms based on training data sets including any, some, or all of the code integration request/outcome data received in steps 1102-1104. In various embodiments, various different types of trained models may be used, including classification systems that execute supervised or semi-supervised learning techniques, such as a Naïve Bayes model, a Decision Tree model, a Logistic Regression model, or a Deep Learning Model, or any other machine learning or artificial intelligence based prediction system that may execute supervised or unsupervised learning techniques").
Regarding claim 3, the combination of Polleri and Gardner teaches The method as in claim 1, wherein the one or more inferences about the artificial intelligence model comprises an inference that the artificial intelligence model fails to meet required generality or robustness, and wherein the replacement artificial intelligence model is configured to meet the required generality or robustness(Polleri [¶0047] "A monitoring engine 156 can monitor operation of the machine learning applications 112 according to the KPI/QoS metrics 160 to assure the machine learning application 112 is performing according to requirements. In addition the monitoring engine 156 can seamlessly test end-to-end a new or evolving machine learning application at different scales, settings, loading, settings, etc" [¶0095] "the machine learning platform can inform the user of the monitored values, and alert the user if the QoS/KPI metrics fall outside prescribed thresholds.").
Regarding claim 4, the combination of Polleri and Gardner teaches The method as in claim 1, wherein the one or more inferences about the artificial intelligence model comprises an inference that it lacks accuracy, (Polleri [¶0102] "At 330, the functionality includes capturing anomalies before they manifest into inaccurate predictions […] the functionality includes fixing inaccurate data automatically or semi-autonomously. In various embodiments, the monitoring engine can determine that received data may be inaccurate. In various embodiments, the monitoring engine can notify a user that the data may be inaccurate")
and wherein the replacement artificial intelligence model has greater accuracy than that of the artificial intelligence model.(Polleri [¶0014] "The machine learning application with the new model can be tested off-line with the results being compared with ground truth data. If the machine learning application with the new model outperforms the previously used model, the machine learning application can be upgraded and auto-promoted to production. One or more parameters may also be discovered. The new parameters may be incorporated into the existing model in an off-line mode. The machine learning application with the new parameters can be tested off-line and the results can be compared with previous results with existing parameters. If the new parameters outperform the existing parameters as compared with ground-truth data, the machine learning application can be auto-promoted to production").
Regarding claim 7, the combination of Polleri and Gardner teaches The method as in claim 1, wherein the data regarding generation of the artificial intelligence model indicates a training environment in which the artificial intelligence model was trained, and wherein the one or more concepts relate to that training environment.(Polleri [¶0105] "A model training system 1016 may retrieve data from data stores 1030 and/or client systems 1050, in order to train models 115 to generate predictive outcomes" [¶0282] "at 1106 the models may be trained using machine-learning algorithms based on training data sets including any, some, or all of the code integration request/outcome data received in steps 1102-1104" [¶0280] " the code integration request data may include one or more data sets corresponding to previous requests made by developers to integrate external code bases into software projects/components. For example, the code integration request data retrieved at 1102 may include the particular characteristics for each of a plurality of code integration requests and the corresponding responses. As noted above, such request characteristics may include, for example [...] the computing and networking environments into which the software is to be deployed" Polleri explicitly states that the model may be trained using code integration request data which is explicitly anticipated as including the computing/networking environment on which the training is run).
Regarding claim 10, the combination of Polleri and Gardner teaches The method as in claim 1, wherein the data regarding generation of the artificial intelligence model comprises one or more class labels used by the artificial intelligence model. (Polleri [[¶0306] "At 1406, the functionality includes extracting one or more features from the data storage. The data storage can include one or more labels that characterize the data. The techniques can automatically detect equivalent entities for the one or more labels. For example, a feature labelled “location” can also recognize data labels for “address.” In various embodiments, the techniques can review the one or more labels that characterize the data to determine the one or more features from the data." [¶0321] "At 1514, the functionality can include generating a mapping of the one or more labels with the categories of the machine learning model. The mapping can identify a location in the data set for each of the categories of the machine learning model." [¶0326] " the method can include presenting the mapping of the one or more labels with the categories of the machine learning model. The technique can include receiving a second input. The second input can correlate to a label of the one or more labels for the schema of the data with a category of the one or more categories known by the machine learning model").
Regarding claims 11-14 and 17, claims 11-14 and 17 are directed towards a system for performing the method of claims 1-4 and 7, respectively. Therefore, the rejections applied to claims 11-14 and 17 also apply equally to claims 1-4 and 7. Claims 11-14 and 17 also recite additional elements “An apparatus, comprising: a network interface to communicate with a computer network and a training system; a processor coupled to the network interface and configured to execute one or more processes” (Polleri [¶0052] "The model execution engine 108 can execute the machine learning application 112 on infrastructure 128 using one or more the infrastructure interfaces 124. The infrastructure 128 can include one or more processors, one or more memories, and one or more network interfaces, one or more buses and control lines that can be used to generate, test, compile, and deploy a machine learning application 112" [¶0433] "FIG. 22 illustrates an exemplary computer system 2200, in which various embodiments of the present invention may be implemented. The system 2200 may be used to implement any of the computer systems described above. As shown in the figure, computer system 2200 includes a processing unit 2204 that communicates with a number of peripheral subsystems via a bus subsystem 2202”).
Regarding claim 20, claim 20 is directed towards “A tangible, non-transitory, computer-readable medium storing program instructions that cause a device to execute” the method of claim 1. Therefore, the rejection applied to claim 1 also applies to claim 20.
Claims 5 and 15 are rejected under U.S.C. §103 as being unpatentable over the combination of Polleri and Gardner and in further view of Zhang ("A Model of Two Tales: Dual Transfer Learning Framework for Improved Long-tail Item Recommendation", 2021).
Regarding claim 5, the combination of Polleri and Gardner teaches The method as in claim 1.
However, the combination of Polleri and Gardner doesn't explicitly teach further comprising: receiving, at the device and via a user interface, an adjustment to the one or more concepts of the knowledge graph.
Zhang, in the same field of endeavor, teaches The method as in claim 1, further comprising: receiving, at the device and via a user interface, an adjustment to the one or more concepts of the knowledge graph. ([p. 2 §1] "the learned meta-mapping can capture the implicit data augmentation in model-level without changing data distribution – for example, given an item with a user feedback, the meta-mapping learns to implicitly add similar user who could give feedback to the item. Therefore, the learned meta-mapping can be leveraged to enhance the representation learning of tail items that contains few user feedback").
The combination of Polleri and Gardner as well as Zhang are directed towards meta-learning. Therefore, the combination of Polleri and Gardner as well as Zhang are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of the combination of Polleri and Gardner with the teachings of Zhang by using user feedback to make adjustment to knowledge graph concepts. Zhang provides as additional motivation for combination ([Abstract] "To improve tail-item recommendation, we conduct research to transfer knowledge from head items to tail items, leveraging the rich user feedback in head items and the semantic connections between head and tail item").
Regarding claim 15, claim 15 is directed towards a system for performing the method of claim 5. Therefore, the rejection applied to claim 5 also applies to claim 15.
Claims 6 and 16 are rejected under U.S.C. §103 as being unpatentable over the combination of Polleri and Gardner and in further view of Rocktaschel (“End-to-End Differentiable Proving”, 2017).
Regarding claim 6, the combination of Polleri and Gardner teaches The method as in claim 1.
However, the combination of Polleri and Gardner doesn't explicitly teach, wherein causing generation of the replacement artificial intelligence model for the artificial intelligence model comprises: converting a symbolic reasoning task of the artificial intelligence model into a deep learning task in the replacement artificial intelligence model.
Rocktaschel, in the same field of endeavor, teaches The method as in claim 1, wherein causing generation of the replacement artificial intelligence model for the artificial intelligence model comprises: converting a symbolic reasoning task of the artificial intelligence model into a deep learning task in the replacement artificial intelligence model. ([Abstract] "we replace symbolic unification with a differentiable computation on vector representations of symbols using a radial basis function kernel, thereby combining symbolic reasoning with learning subsymbolic vector representations. By using gradient descent, the resulting neural network can be trained to infer facts from a given incomplete knowledge base.").
Polleri's platform is about semantic, context-aware composition using ontology semantics and a reasoner engine. Gardner teaches that adding sub-symbolic semantic distances to a knowledge graph helps constrain/search more effectively because the KG includes both logical relationships and contextual semantic distances "in the same graph". Rocktaschel teaches a reasoning method that explicitly uses subsymbolic vectors during symbolic inference, which is naturally compatible with a knowledge graph that already encodes subsymbolic distance information. Therefore, the combination of Polleri and Gardner as well as Rocktaschel are reasonably pertinent analogous art. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of the combination of Polleri and Gardner with the teachings of Rocktaschel by using the symbolic inference model in Rocktaschel as the reasoner engine in Polleri. This combination is a routine implementation choice to make Polleri's existing ontology reasoner more robust. Rocktaschel provides as additional motivation for combination ([p. 9 §8] “We proposed an end-to-end differentiable prover for automated KB completion that operates on subsymbolic representations. To this end, we used Prolog’s backward chaining algorithm as a recipe for recursively constructing neural networks that can be used to prove queries to a KB. Specifically, we introduced a differentiable unification operation between vector representations of symbols. The constructed neural network allowed us to compute the gradient of proof successes with respect to vector representations of symbols, and thus enabled us to train subsymbolic representations end-to end from facts in a KB, and to induce function-free first-order logic rules using gradient descent. On benchmark KBs, our model outperformed ComplEx, a state-of-the-art neural link prediction model, on three out of four KBs while at the same time inducing interpretable rules”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 16, claim 16 is directed towards a system for performing the method of claim 6. Therefore, the rejection applied to claim 6 also applies to claim 16.
Claims 8 and 18 are rejected under U.S.C. §103 as being unpatentable over the combination of Polleri and Gardner and in further view of Dalli (US20220172050A1).
Regarding claim 8, the combination of Polleri and Gardner teaches The method as in claim 1.
However, the combination of Polleri and Gardner doesn't explicitly teach, wherein the replacement artificial intelligence model has fewer total nodes, layers, or nodes per layer than that of the artificial intelligence model..
Dalli, in the same field of endeavor, teaches the replacement artificial intelligence model has fewer total nodes, layers, or nodes per layer than that of the artificial intelligence model. ([¶0210] "a combination of quantization and pruning methods may be applied during the XAED and/or XGAN processing to increase performance and possibly reduce implementation size" [¶0217] "This allows the AutoXAI system to adapt the XAED and/or XGAN system performance to one or more specific application domains or tasks and provides a practical solution to the incorporation of meta-learning systems within an XAED and/or XGAN system, which while common in reinforcement learning systems is an unobvious use for XAEDs/XGANs").
Polleri and Gardner are directed towards meta-learning and Dalli is directed towards explainable AI which incorporates meta-learning. Therefore, Polleri and Gardner as well as Dalli are reasonably pertinent analogous art. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Polleri and Gardner with the teachings of Dalli by pruning the base learner in Polleri and Gardner and incorporating the system on a computer system with a processor. While the benefits of pruning a model would be obvious to one of ordinary skill in the art, they are explicitly reinforced by Dalli who provides as additional motivation for combination that the pruning is (“to increase performance and possibly reduce implementation size” [¶0210]).
Regarding claim 18, claim 18 is directed towards a system for performing the method of claim 8. Therefore, the rejection applied to claim 8 also applies to claim 18.
Claims 9 and 19 are rejected under U.S.C. §103 as being unpatentable over the combination of Polleri and Gardner and Bui (US11556826B2).
Regarding claim 9, the combination of Polleri and Gardner teaches The method as in claim 1, wherein the data regarding generation of the artificial intelligence model comprises information regarding training data used to train the artificial intelligence model, (Polleri [¶0105] "A model training system 1016 may retrieve data from data stores 1030 and/or client systems 1050, in order to train models 115 to generate predictive outcomes").
However, the combination of Polleri and Gardner doesn't explicitly teach and wherein the one or more inferences indicate that the training data should be altered when generating the replacement artificial intelligence model..
Bui, in the same field of endeavor, teaches and wherein the one or more inferences indicate that the training data should be altered when generating the replacement artificial intelligence model.([Col. 15 l. 9-20] "the hyper-parameter determination system 102 searches an enriched hyper-parameter space which augments the original space by including training set size as a hyper-parameter, with candidates selected from varying fractions of a full set size (e.g., 20%, 40%, 60%, 80%, and 100%). In these or other embodiments, the hyper-parameter determination system 102 selects values for k and the number of iterations S. For example, the hyper-parameter determination system 102 selects S=20 and k=3 to balance between accuracy and training efficiency in some cases.").
The combination of Polleri and Gardner as well as Bui are directed towards machine learning hyper-parameter optimization. Therefore, The combination of Polleri and Gardner as well as Bui are reasonably pertinent analogous art. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of The combination of Polleri and Gardner with the teachings of Bui by augmenting the training data responsive to training. Bui recognizes the training set size as one of the hyperparameters to be optimized and provides as additional motivation for combination ([Col. 5 l. 25-42] "the hyper-parameter determination system can improve flexibility for a wider range of real-world applications than many conventional hyper-parameter selection systems. As a result of their accuracy-focused hyper-parameter selection, many conventional systems may not meet the strict real-world efficiency requirements necessary to deploy in a production environment. In addition, many conventional systems focus on optimizing hyper-parameter determination system of a given model class, while ignoring other important extrinsic hyper-parameters (e.g., training set size). The hyper-parameter determination system, on the other hand, can account for extrinsic hyper-parameters within an enriched hyper-parameter space and also improve efficiency over conventional systems for better deployment in real-world applications").
Regarding claim 19, claim 19 is directed towards a system for performing the method of claim 9. Therefore, the rejection applied to claim 9 also applies to claim 19.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SIDNEY VINCENT BOSTWICK whose telephone number is (571)272-4720. The examiner can normally be reached M-F 7:30am-5:00pm EST.
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, Miranda Huang can be reached on (571)270-7092. 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.
/SIDNEY VINCENT BOSTWICK/Examiner, Art Unit 2124