DETAILED ACTION
Claims 1-20 are presented for examination
This office action is in response to submission of application on 13-FEBURARY-2023.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
The amendment filed on 20-JANURARY-2026 in response to the non-final office action mailed 23-OCTOBER-2025 has been entered. Claims 1-20 remain pending in the application.
With regards to the 103 rejections, the applicant’s amendments to the claims have not overcome the rejections to claims 1-20 as the former prior art sufficiently teaches the newly added limitations of the amended claims.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-3, 8, 11-13, 18, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over GANTI (U.S. Pub. No. US 20200005191 A1) in view of SETHI (U.S. Pub. No. US 20240070304 A1) in view of SCHADEWALDT (U.S. Pub. No. US 20210161508 A1)
Regarding claim 1, GANTI teaches the invention as substantially claimed including:
An apparatus comprising at least one processor and at least one non-transitory memory comprising program code stored thereon, wherein the at least one non-transitory memory and the program code are configured to, with the at least one processor, cause the apparatus to at least: receive model configuration data and operational system context data, wherein the model configuration data identifies one or more machine learning models and a respective model training pipeline associated with each of the one or more machine learning models, and the operational system context data identifies one or more objects associated with one or more operational systems, ([0007] According to one or more embodiments of the present invention, computer-implemented methods for ranking and updating machine learning models based on data inputs at edge nodes in a distributed computer system are provided. A non-limiting example computer-implemented method includes receiving, by a processor, an input dataset for training a new machine learning model. For each of a plurality of trained machine learning models, a hash function and a sketch of a training dataset used to train the machine learning model are retrieved. (i.e. model configuration data) A sketch of the input dataset is computed based on the hash function and the input dataset, (i.e. model association data) along with a distance a distance between the sketch of the training dataset and the sketch of the input dataset.(i.e. metadata) [0058] In accordance with one or more embodiments of the present invention, each trained machine learning model (M) that is a candidate seed model is stored along with its corresponding deep hash code generator (H) and semantic label (S). The hash code generator (H), maps the input data to a n-dimensional real-valued vector. The hash code generator (H) is trained on the same dataset that is used to train the machine learning model (M). Unlike the machine learning model (M) whose intent is to capture application semantics (e.g., object localization), the goal of the hash code generator (H) [0066] Drawing an analogy to latent semantic indexing on text documents, the model (M) can be viewed as the text document and its semantic label (S) can be viewed as its index. From a semantic standpoint, deep hash codes determine in a quantitative way the extent to which a given data belongs to the model's training data distribution. (the hash code generated by the hash function is a representation of the model and its data for the training, i.e. configuration data))
While GANTI, does teach metadata relating to operational data, it does not explicitly teach:
and an ontology model describing one or more associations between various components of the one or more operational systems:
However, in analogous art that similarly employs a model, SCHADEWALDT teaches:
and an ontology model describing one or more associations between various components of the one or more operational systems. ([0097] The registration of the captured and saved data to the geometric and semantic model can mean the localization and orientation of the data in the geometric model and the association of the acquired data with the semantic ontology model. This is performed as follows:
[0098] Based on captured labelled measurements, a text-based association to the semantic model is possible, which allows an association with the geometric model. This form of computer-implemented association is known in the art.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with SCHADEWALDT‘s ontological model and, with GANTI‘s metadata, with a reasonable expectation of success, an ontology model, as in SCHADEWALDT, that describes the link between the metadata and the system, as found in GANTI. A person of ordinary skill would have been motivated to improve performance (SCHADEWALDT [0015]).
GANTI further teaches:
generate model association metadata based at least in part on the model configuration data and the operational system context data, wherein the model association metadata defines associations between the one or more machine learning models and the one or more objects, and comprises, for each association, model training parameters and model deployment parameters; ([0007] … along with a distance a distance between the sketch of the training dataset and the sketch of the input dataset.(i.e. metadata) The plurality of trained machine learning models are ranked from smallest computed distance to largest computed distance. [0057] Turning now to an overview of the aspects of the invention, one or more embodiments of the present invention address the above-described shortcomings of the prior art by providing a computer-implemented method and system to search and rank a catalog of pre-trained machine learning models to select a trained machine learning model to be used as a seed model for customizing and re-training based on a new input dataset. The seed model is selected based on a similarity between the training data used to generate the seed model and the data in the new input dataset. (the ranking is meta data which is based on metadata made up from the model association data and model config data)) for each association between a particular machine learning model and a particular object of the associations defined by the model association metadata, train the particular machine learning model according to the model training pipeline associated with the particular machine learning model in the model configuration data based at least in part on operational data associated with the particular object to produce a trained instance corresponding to the association ([0007] The training process is based at least in part on the selected seed machine learning model and the input dataset. [0010] In addition to one or more of the features described above or below, or as an alternative, further embodiments of the computer-implemented method may include initiating re-training of the selected seed machine learning model, the re-training based at least in part on a training dataset used to train the selected seed machine learning model and the input dataset. Technical benefits and advantages can include keeping the machine learning model current and continuously adapting over time. (the training pipeline is chosen based off the metadata and the configuration data and association data by choosing a seed model to base the training process off of. Re-training the seed model generates a trained instance of the seed model.)))
While GANTI, as modified by SCHADEWALDT, does teach training a model based on metadata, association data, and configuration data, it does not explicitly teach:
receive model association input via a model association interface, the model association interface being configured to present a hierarchy of the one or more objects based at least in part on the ontology model, and wherein generating the model association metadata is based at least in part on the model association input… for each trained instance of a trained machine learning model, generate trained model metadata associated with the trained instance based at least in part on the model association metadata, the trained model metadata comprising at least one of: an indication of whether the trained instance is to be deployed, an execution schedule for the trained instance, one or more execution types for the trained instance comprises at least real-time execution or batch execution, and one or more deployment endpoints for the trained instance; and for each trained machine learning model, register the trained machine learning model in a trained model registry, including storing a trained model artifact representing the trained machine learning model and the trained model metadata associated with the trained machine learning model in a data repository associated with the trained model registry; wherein, for each trained instance, the trained model artifact and the trained model metadata corresponding to the trained instance are stored in the data repository associated with the trained model registry.
However, in analogous art that similarly trains a model, SETHI teaches:
receive model association input via a model association interface, the model association interface being configured to present a hierarchy of the one or more objects based at least in part on the ontology model, and wherein generating the model association metadata is based at least in part on the model association input ([0054] Besides the machine learning lifecycle indicator 216, the custom metrics visualization tool 218, and the model performance indicator 220, the user interface 112 is also configured to display, to reviewing users, other metadata about the machine learning model being managed by the model registry 102. For example, the user interface 112 displays the name, the description, the creator's name, the code location, and the data location of the model. In other examples, the user interface 112 displays the number of models that are grouped under a common project.) for each trained instance of a trained machine learning model, generate trained model metadata associated with the trained instance based at least in part on the model association metadata, ([0042] In some embodiments, the machine learning training system 212 is also configured to experiment with different machine learning models on the training and test data. For instance, since the key requirement when the machine learning training system 212 experiments with different machine learning models is to enable the users to choose the best model, metadata captured in this phase of experimentation includes items such as performance metrics for the model, hyperparameter values used during training, lifecycle indicators of the model, etc.) , the trained model metadata comprising at least one of: an indication of whether the trained instance is to be deployed, an execution schedule for the trained instance, one or more execution types for the trained instance comprises at least real-time execution or batch execution, and one or more deployment endpoints for the trained instance; ([0066] In some example implementations, the container model use case program 308 is also configured to deploy the retrieved machine learning model via the model management API 310 in step 412. Sitting in a system external to where machine learning models are stored, the model management API 310 is used as an interface to pass in data required for running a machine learning application. The user can access a data link that is linked to the model management API 310 and thereby get outputs of the machine learning application through the model management API 310. By deploying the model through the model management API 310, the user of the model does not need to manage all the background processes involving the model. Therefore, the usability of the machine learning model is significantly improved.) and for each trained machine learning model, register the trained machine learning model in a trained model registry, including storing a trained model artifact representing the trained machine learning model and the trained model metadata associated with the trained machine learning model in a data repository associated with the trained model registry; [0043] After training the machine learning model, the machine learning training system 212 is then configured to send data of the trained model and the captured metadata about the trained model to the model registry 102 operating on, for example, the computing system 120 (illustrated and described in FIG. 1). In some embodiments, users of the model registry 102 identify and retrieve the desired model base on the stored metadata capturing and describing the model. In other embodiments, users of the model registry 102 retrieve the data of the model itself.) wherein, for each trained instance, the trained model artifact and the trained model metadata corresponding to the trained instance are stored in the data repository associated with the trained model registry. ([0068] In some example implementations, the model management API 310 is further configured to send the metadata and metrics of the machine learning model, which the model management API 310 receives from the model registry 102, to the relational database management system 314. By storing the metadata and metrics of the machine learning model with the relational database management system 314, more advanced operations on the metadata and metrics are enabled. For instance, the relational database management system 314 can be a POSTGRES database management system compatible with Structured Query Language (SQL). The relational database management system 314 is used for data analysis and creating reports related to the metadata and metrics of the machine learning model.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with SETHI‘s registry storage and, with GANTI‘s, as modified by SCHADEWALDT, model data and metadata and trained model, with a reasonable expectation of success, a storage registry, as in SETHI, that stores trained models, as found in GANTI, as modified by SCHADEWALDT. A person of ordinary skill would have been motivated to decrease retraining time (SETHI [0005]).
Regarding claim 2, GANTI further teaches:
The apparatus of claim 1, wherein the model association metadata comprises the model training parameters and the model deployment parameters corresponding to each association between a particular machine learning model and a particular object of the associations defined by the model association metadata. ([0058] In accordance with one or more embodiments of the present invention, each trained machine learning model (M) that is a candidate seed model is stored along with its corresponding deep hash code generator (H) and semantic label (S). The hash code generator (H), maps the input data to a n-dimensional real-valued vector. The hash code generator (H) is trained on the same dataset that is used to train the machine learning model (M). Unlike the machine learning model (M) whose intent is to capture application semantics (e.g., object localization), the goal of the hash code generator (H) is to capture the characteristics of the training data used to train the machine learning model (M). Semantic label (S) is a compact sketch over the hash codes derived from the training data (e.g., via unsupervised clustering). As used herein, the terms “sketch over the hash codes” refers to creating a compact representation (e.g., a cluster head; a function capturing the hash codes) of the hash codes so that all the hash codes are not stored. [0059] In accordance with one or more embodiments of the present invention, a protocol is provided between a user (e.g., at an edge node) and a cloud hosted catalog of machine learning models that allow the user to search using a given input dataset. Search results returned include a list of machine learning models in the catalog that are ranked based on a semantic distance measure between the search dataset and the training dataset used to train the machine learning model (M). In accordance with one or more embodiments of the present invention, the machine learning model having the highest rank (e.g., lowest distance) is selected as the seed model for training based on the input dataset. In other embodiments of the present invention, a machine learning model ranked within the top two or three or some other threshold number is selected as the seed model.)
Regarding claim 3, SETHI further teaches:
The apparatus of claim 1, wherein each trained machine learning model registered in the trained model registry is executed based at least in part on the stored trained model artifact and trained model metadata corresponding to the trained machine learning model. ([0066] In some example implementations, the container model use case program 308 is also configured to deploy the retrieved machine learning model via the model management API 310 in step 412. Sitting in a system external to where machine learning models are stored, the model management API 310 is used as an interface to pass in data required for running a machine learning application. The user can access a data link that is linked to the model management API 310 and thereby get outputs of the machine learning application through the model management API 310. By deploying the model through the model management API 310, the user of the model does not need to manage all the background processes involving the model. Therefore, the usability of the machine learning model is significantly improved.
[0067] The model management API 310 is configured to send the metadata and metrics of the machine learning model, which the model management API 310 receives from the model registry 102, to the user interface 312 in step 414. The user interface 312 is then configured to display the metadata and metrics for the user to review.)
Regarding claim 8, GANTI further teaches:
The apparatus of claim 1, wherein the operational system context data comprises at least one of: metadata associated with various components of the one or more operational systems ([0007] According to one or more embodiments of the present invention, computer-implemented methods for ranking and updating machine learning models based on data inputs at edge nodes in a distributed computer system are provided. A non-limiting example computer-implemented method includes receiving, by a processor, an input dataset for training a new machine learning model. For each of a plurality of trained machine learning models, a hash function and a sketch of a training dataset used to train the machine learning model are retrieved. A sketch of the input dataset is computed based on the hash function and the input dataset, along with a distance a distance between the sketch of the training dataset and the sketch of the input dataset. The plurality of trained machine learning models are ranked from smallest computed distance to largest computed distance. A seed machine learning model is selected from the plurality of trained machine learning models based at least in part on the ranking. A training process of the new machine learning model is initiated based at least in part on the selecting. The training process is based at least in part on the selected seed machine learning model and the input dataset.)
SCHADEWALDT further teaches:
and an ontology model describing one or more associations between various components of the one or more operational systems. ([0097] The registration of the captured and saved data to the geometric and semantic model can mean the localization and orientation of the data in the geometric model and the association of the acquired data with the semantic ontology model. This is performed as follows:
[0098] Based on captured labelled measurements, a text-based association to the semantic model is possible, which allows an association with the geometric model. This form of computer-implemented association is known in the art.)
Regarding claims 11-13, they comprise of limitations similar to those of claims 1-3 and are therefore rejected for similar rationale. Regarding claim 18, it comprises of limitations similar to those of claim 8 and is therefore rejected for similar rationale. Regarding claim 20, it comprises of limitations similar to those of claim 1 and is therefore rejected for similar rationale.
Claims 4 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over GANTI (U.S. Pub. No. US 20200005191 A1), SETHI (U.S. Pub. No. US 20240070304 A1), SCHADEWALDT (U.S. Pub. No. US 20210161508 A1) in further view of AGUILAR (U.S. Pub. No. US 20220188704 A1)
Regarding claim 4, SETHI further teaches:
The apparatus of claim 1, wherein the trained model metadata for each trained machine learning model comprises at least one of: an indication of whether the trained machine learning model is to be deployed, … and one or more deployment endpoints for the trained machine learning model. ([0051] The machine learning lifecycle indicator 216 is configured to display an indicator that indicates the current lifecycle phase of the machine learning model being managed by the model registry 102. For instance, the lifecycle of a machine learning model can be divided into five phases, namely: (1) data preparation, that is obtaining the training and test data to develop a model; (2) feature engineering, that is identifying or creating the appropriate descriptors from the input data (i.e., features) to be used by the model; (3) model training and experimentation, that is experimenting with different models on the training and test data and choosing the best; (4) deployment, that is deploying the chosen model in an inferencing system; and (5) maintenance, that is monitoring the model performance, updating the model as needed, and eventually retiring the model.)
While SETHI does teach deploying a model with endpoints, it does not explicitly teach:
an execution schedule for the trained machine learning model, one or more execution types for the trained machine learning model
However, in analogous art that similarly uses a model, AGUILAR teaches:
an execution schedule for the trained machine learning model, ([0012] The ML model metadata may further comprise scheduling information for that ML model, and the method may further comprise scheduling execution of the ML model using the scheduling information.) one or more execution types for the trained machine learning model, ([0008] The ML model metadata may further comprise information specifying a processing resource type for executing the ML model.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with AUGILAR‘s execution schedule and type and, with GANTI‘s, as modified by SETHI, model data and metadata and model execution, with a reasonable expectation of success, an execution schedule and type, as in AUGILAR, of a model used, as found in GANTI, as modified by SETHI. A person of ordinary skill would have been motivated to better optimize models (AUGILAR [0081]).
Regarding claim 14, it comprises of limitations similar to those of claim 4 and is therefore rejected for similar rationale.
Claims 5 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over GANTI (U.S. Pub. No. US 20200005191 A1), SETHI (U.S. Pub. No. US 20240070304 A1), SCHADEWALDT (U.S. Pub. No. US 20210161508 A1) in further view of RAVID (U.S. Pub. No. US 20180239615 A1)
While GANTI, as modified by SETHI, does teach claim 1, which claim 5 is dependent upon, it does not explicitly teach:
The apparatus of claim 1, wherein the trained model metadata for each trained machine learning model comprises one or more execution types for the trained machine learning model, the one or more execution types including at least one of: real time execution and batch execution.
However, in analogous art that similarly executes a model, RAVID teaches:
The apparatus of claim 1, wherein the trained model metadata for each trained machine learning model comprises one or more execution types for the trained machine learning model, the one or more execution types including at least one of: real time execution and batch execution. ([0136] In this example, CDA system 753 is configured to push (e.g., in batch and/or in real-time) profile data in profile data structure 771 to collect application 780, which in turn transmits the profile data to external input systems 769 and/or networked systems 768. CDA system 753 is also configured to push (e.g., in batch and/or in real-time) profile data from profile data structure 771 to data layer 788 for transmittal to external output systems 770, networked systems 768 as shown by data flow 757 (e.g., to enable operational scheduling, monitoring, audit trails and execution tracking based on data included in the profile data, such as, data specifying what actions were executed and when those actions were execution) and to external input systems 769 as shown by data flow 791 (e.g., to enable data governance and data management, such as, data profiling and quality and data lineage based on metadata and reference data included in the profile data). In this example, profile data stored in profile data structure 771 includes data lineage data and data representing transformations and other operations performed on one or more portions of profile data (e.g., associated with a particular key).)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with RAVID‘s execution types and, with GANTI‘s, as modified by SETHI, models, with a reasonable expectation of success, a method of execution, as in RAVID, applied to a model, as found in GANTI, as modified by SETHI. A person of ordinary skill would have been motivated to improve execution time and consistency (RAVID [0003]).
Regarding claim 15, it comprises of limitations similar to those of claim 5 and is therefore rejected for similar rationale.
Claims 6 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over GANTI (U.S. Pub. No. US 20200005191 A1), SETHI (U.S. Pub. No. US 20240070304 A1), SCHADEWALDT (U.S. Pub. No. US 20210161508 A1) in further view of LEVANON (U.S. Pub. No. US 20020143791 A1)
Regarding claim 6, while GANTI, as modified by SETHI, does teach claim 1, which claim 6 is dependent upon, it does not explicitly teach:
The apparatus of claim 1, wherein the model association metadata is generated based at least in part on model association input received via a model association interface.
However, in analogous art that similarly executes a model, LEVANON teaches:
The apparatus of claim 1, wherein the model association metadata is generated based at least in part on model association input received via a model association interface. ([0056] The interface may be operative to determine the type of pre-generated content and provides a number of fields for input of data in dependence on the type, wherein the interface is operative to generate the metadata in dependence on the input data.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with LEVANON‘s metadata generation method and, with GANTI‘s, as modified by SETHI, model data, with a reasonable expectation of success, a method generating metadata, as in LEVANON, using model data, as found in GANTI, as modified by SETHI. A person of ordinary skill would have been motivated to reduce complexity (LEVANON [0009]).
Regarding claim 16, it comprises of limitations similar to those of claim 6 and is therefore rejected for similar rationale.
Claims 7 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over GANTI (U.S. Pub. No. US 20200005191 A1), SETHI (U.S. Pub. No. US 20240070304 A1), SCHADEWALDT (U.S. Pub. No. US 20210161508 A1) in further view of STUMP (U.S. Pub. No. US 20230058094 A1) in further view of SUMENKOV (E.P. Pub. No. EP 3674943 A1)
Regarding claim 1, while GANTI, as modified by SETHI, does teach claim 1, which claim 7 is dependent upon, it does not explicitly teach:
The apparatus of claim 1, wherein the one or more objects associated with the one or more operational systems include at least one of: one or more assets of the one or more operational systems, …and one or more alarms defined for the one or more operational systems.
However, in analogous art that similarly uses a model, STUMP teaches:
The apparatus of claim 1, wherein the one or more objects associated with the one or more operational systems include at least one of: one or more assets of the one or more operational systems, …and one or more alarms defined for the one or more operational systems. ((STUMP claim 10.) The system of claim 1, wherein the edit modifies an attribute of the automation object, the attribute comprising least one of control code for monitoring and controlling an industrial asset represented by the automation object, a visualization object that defines a graphical visualization of the industrial asset, an alarm definition for the industrial asset, a security feature of the industrial asset, a security protocol of the industrial asset, a test script configured to validate operation of the automation object within the system project, or an analytic script configured to perform an analysis on data generated by the industrial asset.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with STUMP‘s assets and alarms and, with GANTI‘s, as modified by SETHI, system, with a reasonable expectation of success, objects that include assets and alarms, as in STUMP, that belong to a system, as found in GANTI, as modified by SETHI. A person of ordinary skill would have been motivated to improve performance (STUMP [0120]).
While GANTI, as modified by SETHI and STUMP, does teach a system with objects made up of assets and alarms, it does not explicitly teach:
one or more sites containing the one or more operational systems,
However, in analogous art that similarly employs a ML model, SUMENKOV teaches:
one or more sites containing the one or more operational systems, ([0026] In one example, the objects 111 may comprise at least: files; network packets; sites; RAM pages, both physical and virtual; processes and other objects of the operating system related to them; SUMENKOV )
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with SUMENKOV‘s sites and, with GANTI‘s, as modified by SETHI and STUMP, system with asset and alarm objects, with a reasonable expectation of success, objects that are included in sites, as in SUMENKOV, that belong to a system, as found in GANTI, as modified by SETHI and STUMP. A person of ordinary skill would have been motivated to improve efficiency (SUMENKOV [0007])
Regarding claim 17, it comprises of limitations similar to those of claim 7 and is therefore rejected for similar rationale.
Claims 9-10 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over GANTI (U.S. Pub. No. US 20200005191 A1), SETHI (U.S. Pub. No. US 20240070304 A1), SCHADEWALDT (U.S. Pub. No. US 20210161508 A1) in further view of LLORENTE (U.S. Pub. No. US 20240356815 A1)
While GANTI, as modified by SETHI, does teach claim 1, which claim 9 is dependent upon, it does not explicitly teach:
The apparatus of claim 1, wherein the at least one non-transitory memory and the program code are configured to, with the at least one processor, further cause the apparatus to at least: monitor execution of each trained machine learning model registered in the trained model registry based at least in part on predefined monitoring criteria.
However, in analogous art that similarly applies a model, LLORENTE teaches:
The apparatus of claim 1, wherein the at least one non-transitory memory and the program code are configured to, with the at least one processor, further cause the apparatus to at least: monitor execution of each trained machine learning model registered in the trained model registry based at least in part on predefined monitoring criteria. ([0512] B9. The method of embodiment B8, further comprising, at each monitoring period: [0513] determining respective first relations between the respective metrics for the monitoring objects and respective thresholds associated with the monitoring objects; [0514] determining whether drift has occurred based on a second relation among the respective first relations; and [0515] when it is determined that drift has occurred, determining whether the drift is severe based on a termination threshold.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with LLORENTE‘s execution monitoring and, with GANTI‘s, as modified by SETHI, model to be executed, with a reasonable expectation of success, a method of monitoring a model’s execution, as in LLORENTE, using a trained model, as found in GANTI, as modified by SETHI. A person of ordinary skill would have been motivated to improve accuracy (LLORENTE [0107]).
Regarding claim 10, LLORENTE further teaches:
The apparatus of claim 9, wherein the monitoring of the execution of each trained machine learning model registered in the trained model registry comprises, in response to detecting model drift associated with a trained machine learning model exceeding a predefined drift threshold of the predefined monitoring criteria based at least in part on the execution of the trained machine learning model, at least one of: generating a model drift notification associated with the trained machine learning model and triggering retraining of the trained machine learning model. ([0512] B9. The method of embodiment B8, further comprising, at each monitoring period: [0513] determining respective first relations between the respective metrics for the monitoring objects and respective thresholds associated with the monitoring objects; [0514] determining whether drift has occurred based on a second relation among the respective first relations; and [0515] when it is determined that drift has occurred, determining whether the drift is severe based on a termination threshold.
[0516] B10. The method of any of embodiments B4-B6, wherein: [0517] each drift monitoring notification is received based on a determination by the DDLF that drift has occurred; and [0518] each drift monitoring notification includes one or more of the following: [0519] an indication of whether the determined drift is severe; [0520] an identifier of the ML model or of analytics associated with the drift monitoring; [0521] a timestamp; [0522] a periodicity of the drift monitoring; and [0523] a value of a metric for the type of data associated with each monitoring object.
[0524] B11. The method of any of embodiments B1-B10, further comprising, based on an indication by the DDLF or a determination by the MTLF that the drift of the ML model has occurred, retraining the ML model and notifying the AnLF of availability of a retrained ML model. )
Regarding claim 19, it comprises of limitations similar to those of claim 9 and is therefore rejected for similar rationale.
Response to Arguments
Applicant’s arguments filed 20-JANURARY-2026 have been fully considered, but they are found to be non-persuasive
With regards to the applicant’s remarks regarding the 103 rejection in the non-final action, the applicant argues that the prior art does not teach the newly amended claims 1, 11, and 20. The examiner acknowledges this argument and has adjusted the prior art of GANTI and SETHI to disclose the newly added limitations while adding SCHADEWALDT to claim 1’s mapping. Further, the examiner has adjusted all dependent claims accordingly.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SKIELER A KOWALIK whose telephone number is (571)272-1850. The examiner can normally be reached 8-5.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mariela D Reyes can be reached at (571)270-1006. 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.
/SKIELER ALEXANDER KOWALIK/Examiner, Art Unit 2142
/Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142
Prosecution Insights