DETAILED ACTION
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/Arguments
Applicant’s arguments filed on February 19, 2026 regarding the rejection under 35 U.S.C. 103 have been fully considered but are not persuasive.
Applicant’s arguments presented on pages 18-21 have been considered but are not persuasive. Applicant argues that the cited references fail to teach or suggest “retrieving the AI model template in the transportable container from a templates store at a remote site” and further fails to teach or suggest “orchestrating a lifecycle of an AI model at a local site” including instantiating, training, deploying, monitoring, and re-training the AI model at the local site. Applicant additionally argues that Sharma is deficient because Sharma describes creation and training of a machine-learning model in a remote network rather than a local site. Applicant further argues that the cited references fail to teach or suggest “deploying the AI model as a service to generate one or more predictions and corresponding explanations at the local site,” and specifically argues that the cited references fail to teach the claimed “corresponding explanations.”
Applicant’s arguments are not persuasive because they improperly import limitations from exemplary embodiments in the specification into the claims. While Applicant cites portions of the specification describing embodiments in which AI models are trained on a client site “with no need to transport data outside the client site” and continuously adapt locally to changing data (Spec, paragraphs [0051] and [0055]), the claims do not recite that all lifecycle operations occur exclusively locally, without cloud interaction, or without participation of remote systems. Instead, the claims expressly begin by reciting “retrieving the AI model template in the transportable container from a templates store at a remote site,” thereby expressly contemplating a distributed architecture involving both remote and local computing environments. Thus, the claims themselves contemplate interaction between remote systems and localized execution environments rather than an exclusively isolated local-only implementation.
The claims are interpreted under the broadest reasonable interpretation in light of the specification, but limitations from preferred embodiments are not imported into the claims. Under the broadest reasonable interpretation, the claimed “local site” reasonably encompasses localized execution environments such as Sharma’s edge platform and Polleri’s client infrastructure and local data storage architecture.
As set forth in the rejection, Polleri teaches deployment of machine-learning applications to client infrastructure, local data storage electrically connected to processing circuitry used to generate, test, and execute the application, local retrieval and used of customer data, iterative training and serving cycles using performance metrics, and deployment of machine-learning applications as service to client infrastructure. Sharma further teaches inferencing generated at the edge platform, analytics evaluation on the edge platform, monitoring for drift or degradation of model accuracy at the edge platform, updating and replacing machine-learning models on the edge computing platform in response to identified drift, and generation of “predictions, inferences, analytical results, and other intelligence information, e.g., business and operational insights.” Under the broadest reasonable interpretation, the claimed “corresponding explanations” broadly encompass contextual, analytical, or interpretive information associated with predictions reasonably correspond to the claimed corresponding explanations. Collectively, the cited references teach or at least reasonably suggest instantiating, training, deploying, monitoring, and updating machine-learning models at a localized environment while operating within a distributed architecture that includes remote systems.
Applicant’s arguments are additionally not commensurate with the scope of the claims because the claims do not exclude remote model creation, remote template repositories, distributed model lifecycle architectures, remote-assisted updates, or hybrid edge/cloud machine-learning environments. Further, the claims do not recite any particular explainable artificial intelligence technique, saliency analysis, feature attribution methodology, causal inference mechanism, or specialized explanatory output structure. Applicant’s arguments additionally focus on isolated portions of individual references rather than addressing the combined teachings of the cited references as relied upon in the rejection. Obviousness is determined based on the collective teachings and suggestions of the combined references, not whether any single reference independently discloses every claimed feature in isolation.
Applicant’s arguments regarding claims 2-6, 8-13, and 15-20 have been fully considered but are not persuasive. Applicant does not present separate substantive arguments for these claims beyond those addressed above with respect to claim 1. Claims 2-6 depend from claim 1, claims 9-13 depend from claim 8, and claims 16-20 depend from claim 15. Further, claims 8 and 15 recite limitations substantially similar to those discussed above with respect to claim 1, but in computer program product and computer system formats, respectively. Accordingly, for at least the reasons discussed above with respect to claim 1, the combination of the cited references teaches or at least reasonably suggest the limitations of claims 2-20.
Applicant additionally references claims 21 and 22 in the remarks. However, claims 21 and 22 are not presently pending and therefore not addressed herein.
Applicant’s arguments regarding claims 7 and 14 have been fully considered but are not persuasive. Applicant first argues that Zhao does not cure the alleged deficiencies of Polleri and Sharma with respect to independent claims 1, 8, and 15. This argument is not persuasive for the reasons discussed above with respect to independent claim 1. Claims 7 and 14 depend from claims 1 and 8, respectively, and Applicant’s arguments regarding the base claims have already been addressed.
Applicant further argues the combination of cited references fails to teach or suggest that the AI model template “comprises code, AI model information, an AI model manifest, pipeline metadata, and binary execution images.” This argument is also not persuasive.
As set forth in the rejection for claims 7 and 14, Polleri teaches executable machine-learning application code and associated machine-learning model metadata identifying model types and intended uses. Zhao further teaches that a contributor describes model metadata when publishing, including function descriptions, input/output formats, and model categories. Zhao also teaches that the Acumos platform packages uploaded models into deployable Docker images capable of execution across runtime environments.
Under the broadest reasonable interpretation, Polleri’s executable machine-learning application code and model metadata reasonably correspond to the claimed code and AI model information. Zhao’s metadata, including function descriptions, input/output formats, and model categories, reasonably corresponds to descriptive model information, including an AI model manifest and pipeline metadata associated with the model and its deployment/use. Zhao’s deployable Docker images reasonably correspond to binary execution images and runtime artifacts packaged for transportable deployment.
Applicant’s reliance on Figures 6A and 6C is also not persuasive because the claims do not recite the particular structures, schemas, layouts, or implementation details shown in those exemplary figures. Rather, the claims broadly recite “AI model information,” “AI model manifest,” “pipeline metadata,” and “binary execution images.” Accordingly, under the broadest reasonable interpretation, the combination of Polleri and Zhao teaches or at least reasonably suggests the additional limitations of claims 7 and 14.
Accordingly, for at least the reasons set forth above, the combination of the cited references teaches or at least reasonably suggest the limitations of claims 1-20 under the broadest reasonable interpretation of the pending claims. Therefore, the rejection of claims 1-20 under 35 U.S.C. 103 is maintained.
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 1-6, 8-13, and 15-20 are rejected under the 35 U.S.C. 103 as being unpatentable over Polleri et al., (Pub No.: US 20210081720 A1 (Filed: 2020)) in view of Sharma et al., (Pub. No.: US 20200327371 A1 (Filed: 2019)).
Regarding claim 1, Polleri in view of Sharma teaches the following limitations:
A computer-implemented method, comprising operations for: receiving selection of an Artificial Intelligence (Al) model template at a local site, wherein the AI model template is created at a remote site (Polleri, paragraph [0012] “According to some implementations, a server system may…perform operations to: receive an input, wherein the input identifies a problem to be solved using the machine-learning application…to select a machine-learning model template from a plurality of templates based at least in part on the input, wherein the machine-learning model template includes metadata, the metadata specifies data expectations and available data formats.” Paragraph [0035] “A model composition engine 132… can receive inputs from a user 116 through an interface 104…The model composition engine 132 can interface with library components 168 to identify various pipelines 136, micro service routines 140, software modules 144, and infrastructure models 148 that can be used in the creation of the machine-learning model 112.” – the model template is selected and created at the remote site (server) based on user input (local site). and is packaged in a transportable container (Polleri, paragraph [0036] “ The model composition engine 132 can output one or more machine-learning applications 112. The machine-learning applications 112 can be stored locally on a server or in a cloud-based network. The model composition engine 132 can output the machine-learning application 112 as executable code that be run on various infrastructure 128 through the infrastructure interfaces 124.” Sharma paragraph [0105] “Containers were developed as a part of the popular Linux open-source operating system…Docker containers wrap up a piece of software in a complete file system that contains everything the software needs to run: code, runtime, system tools, and system libraries—anything that can be installed on a server. This ensures the software will always run the same, regardless of the environment in which it is running. Thus, by incorporating a container technology such as Docker in the SDK, applications developed using the SDK can be deployed and executed not only on the edge platform for which they were developed, but also on other edge platform implementations, as well as in the cloud.”)
retrieving the AI model template in the transportable container from the remote site (Polleri [0065] “The selected machine-learning model can be selected for the machine-learning application. The selected machine-learning model can be downloaded and stored locally. In various embodiments, the system can inform the user of the specific model name or identifier of the selected model.” [0098] “Server 412 may be communicatively coupled with remote client computing devices 402, 404, 406, and 408 via network 410.” Sharma [0106] “In the example embodiment, containers for applications are created, deployed, and managed intelligently based on factors such as the types of edge environments and devices involved. This intelligence may be implemented in the form of a software apparatus comprising a number of components including centralized administration, deployment topology templates, container mobility management, zero-touch edge deployment, container monitoring, and responsive migration.” – Polleri teaches retrieving the AI model template from a remote site by disclosing that the selected machine-learning model is downloaded and stored locally. Polleri further describes a distributed network architecture in which server systems and client devices are communicatively coupled via a network. Under BRI, downloading selected model from a remote server system to a local storage reasonably suggests retrieving the AI model template from a remote site. Sharma further teaches that machine-learning applications are packaged and deployed using transportable containers, including “container mobility management” and “zero-touch edge deployment.” These teachings support transporting and deploying containerized model components across distributed environments, which reasonably suggest retrieving the AI model template in a transportable container from a remote site. Accordingly, the combination of Polleri and Sharma teaches or suggests retrieving the AI model template in the transportable container from the remote site.);
orchestrating a lifecycle of an AI model by: instantiating the AI model from the AI model template at the local site (Polleri, paragraph [0036] “The model composition engine 132 can output the machine-learning application 112 as executable code that be run on various infrastructure 128 through the infrastructure interfaces 124.” Paragraph [0037] “The model execution engine 108 can execute the machine-learning application 112 on infrastructure 128 using one or more the infrastructure interfaces 124.” [0044] “The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” [0065] “ The selected machine-learning model can be downloaded and stored locally.” – Polleri teaches instantiating the AI model from the AI model template at a local site. Polleri discloses that the machine-learning application is output as executable code and executed on infrastructure 128 through infrastructure inferences. Polleri further teaches that the data storage location may be local and electrically connected to the processing circuitry used to “generate, test, and execute the application,” and that the selected model can be downloaded and stored locally. Under the broadest reasonable interpretation, executing a locally stored machine-learning application on local infrastructure using locally connected processing circuitry reasonably suggests instantiating an AI model from the AI model template at the local site.);
retrieving data from one or more local data sources at the local site (Polleri, paragraph [0044] “The machine-learning platform 100 can include one or more data storage locations 170. The user can identify the one or more data storage locations 170. The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” [0046] “The model execution engine 108 can use hosted input data 164 to execute and test the machine-learning application 112. The hosted input data 164 can include a portion of the data stored at the data storage 170.” – Polleri teaches retrieving data from one or more local data sources at the local site by teaching that the machine-learning platform includes data storage locations that “can be local” and electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application. Polleri further teaches that the model execution engine uses hosted input data to execute and test the machine-learning application, where the hosted input data includes data stored at the data storage location 170. Under BRI, accessing and using data stored at locally connected data storage locations to execute and test the machine-learning application reasonably suggests retrieving data from one or more local data sources at the local site.)
training the AI model using the data at the local site (Polleri, [0044] “ The machine-learning platform 100 can include one or more data storage locations 170…The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” Paragraph [0054] “At 206, the model composition engine 132 can match one or more of the key features of the selected machine-learning model with a customer data schema of the customer data 208. In this way the customer data 208 can be used with the selected machine-learning model.” Paragraph [0056] “At 214, the model composition engine 132 can capture metrics for the customer data. The model can be trained and validated using known values.” – Polleri teaches using customer data with the selected machine-learning model and further teaches that the model can be trained and validated using known values. Polleri also teaches that the customer data maybe be stored in a local data storage location that is electrically connected to the processing circuitry used to generate, test, and execute the application. Under BRI, training the model using customer data that is stored locally and used by the model composition engine really corresponds to training the AI using the data at the local site. Accordingly, Polleri teaches or suggests training the AI model using the data at the local site.);
deploying the AI model as a service to generate one or more predictions and corresponding explanations at the local site (Polleri paragraph [0094] “ In some implementations, process 300 includes deploying the machine-learning application to a client infrastructure. Service deployment can start with a standard configuration as defined in the machine model template.” Paragraph [0099] “ In various embodiments, server 412 may be adapted to run one or more services or software applications provided by one or more of the components of the system. In some embodiments, these services may be offered as web-based or cloud services or under a Software as a Service (SaaS) model to the users of client computing devices 402, 404, 406, and/or 408.” Sharma, paragraph [0183] “Various predictions, inferences, analytical results, and other intelligence information, e.g., business and operational insights produced by applications, analytics expressions, and machine learning models executing on the edge platform” – Polleri teaches deploying the AI model as a service by describing deployment of the machine-learning application to a client infrastructure and operation of the application as a service available to client devices. Sharma teaches that machine-learning models executing on the edge platform generating predictions together with associated analytical results, intelligence information, and business and operational insights. Under BRI, generating predictions together with such intelligence information reasonably suggests providing accompanying contextual or interpretive information corresponding to the prediction output. Because Sharma locates these outputs on the edge platform, execution on the edge platform corresponds to execution at a local site. Accordingly, the combination of Polleri and Sharma teaches or suggests deploying the AI model as a service to generate predictions and corresponding explanatory information at a local site.);
monitoring the AI model for drift at the local site (Polleri, paragraph [0032] “A monitoring engine 156 can monitor operation of the machine-learning applications 112 according to the key performance indicators (KPI)/Quality of Service (QoS) metrics 160 to assure the machine-learning application 112 is performing according to requirements.” Sharma paragraph [0023] “A model update cycle may be initiated by a trigger, which may comprise a manual trigger, a time-based trigger, or a trigger derived from evaluating inferences generated by the model at the edge. Analytics expressions implementing selected logic, math, statistical or other functions may be applied to a stream of inferences generated by the model on the edge platform. The analytics expressions define what constitutes an unacceptable level of drift or degradation of model accuracy and track the model output to determine if the accuracy has degraded beyond an acceptable limit.” Polleri teaches monitoring operations of a machine-learning application using KPI and QoS metrics to ensure that the application performs according to the requirements. Sharma further teaches evaluating inferences generated at the edge and applying analytics expressions to determine whether the model accuracy has experienced drift or degradation beyond an acceptable limit. Under BRI, evaluating model outputs and determining whether model accuracy has degraded beyond a threshold reasonably corresponds to monitoring the AI model for drift. Because Sharma locates the inferencing, analytics evaluation, and drift determination on the edge platform, executing and monitoring on the edge platform reasonably correspond to monitoring the AI model for drift at the local site. Accordingly, the combination of Polleri and Sharma teaches or suggests monitoring the AI model for drift at the local site.); and
in response to identifying the drift, re-training the AI model at the local site (Sharma, paragraph [0023] “A model update cycle may be initiated by a trigger…The analytics expressions define what constitutes an unacceptable level of drift or degradation of model accuracy …In response, the edge platform can automatically take action…” [0024] “In an implementation of another aspect of the invention, dynamic non-disruptive machine learning model update and replacement on the edge computing platform is provided…a machine learning model on the edge computing platform may be updated with a modified machine learning model…” Polleri, paragraph [0094] “After a few cycles of training and serving, data and performance metrics can be used to automatically adjust the deployment configuration…” – Sharma teaches initiating a model update cycle in response to determining that model accuracy has experienced drift or degradation beyond and acceptable limit. Sharma further teaches updating and replacing the machine-learning model without interrupting ongoing inference operations. Under BRI, updating the model on the edge platform in response to identified drift reasonably corresponds to re-training the AI model at the local site. Polleri additionally teaches iterative cycles of training and serving in which performance metrics drive automatic adjustments to model behavior, reinforcing that model updates may be performed based on observed performance characteristics. Because Sharma expressly locates the inferencing, drift evaluation, and model update operations on the edge computing platform, the edge platform reasonably corresponds to the claimed local site. Accordingly, the combination of Sharma and Polleri teaches or suggests re-training the AI model at the local site in response to identifying drift.)
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having a combination of Polleri and Sharma before them, to incorporate the use of an advanced software container for ensuring transportability, as detailed by Sharma, into the model deployment system described by Polleri. One would have been motivated to make such a combination because Polleri details outputting the machine learning application as executable code that can be run on various infrastructure, and Sharma teaches that this self-contained packaging wraps up the software and its dependencies to ensure it will always run the same, regardless of the environment in which it is running. This capability is necessary to produce a reliable, universally deployable, and easily transportable container that overcomes configuration barriers at local sites.
Regarding claim 2, Polleri in view of Sharma teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Polleri in view of Sharma further teaches:
displaying the AI model template in an AI model template catalog (Polleri, paragraph [0011] “The method can include selecting a machine-learning model template from a plurality of templates based at least in part on the input, wherein the machine-learning model template includes metadata, the metadata specifies data expectations and available data formats.” Sharma [0267] “ The machine learning model software platform presents to the browser a user interface that includes a list of workflows. Each workflow comprises one or more sensor data streams to be operated on and one or more analytics expressions, applications, machine learning models, and others.” [0268] “ A user can select a workflow from the list 1104 to see the details of the various components making up the workflow, which may include one or more machine learning models. A user can edit the parameters associated with a machine learning model that is already part of an existing workflow and also can identify and add a new machine learning model to an existing workflow. For example, a user can enter certain machine learning model information and parameters 1106, such as model type (e.g., regression), algorithm type (e.g., decision tree), and algorithm parameters, using the user interface. These parameters are stored in the software platform.” – functional definition of a catalog (an organized list) and the act of displaying that list in a user interface.).
Regarding claim 3, Polleri in view of Sharma teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Polleri in view of Sharma further teaches:
wherein the AI model is configurable with configurable parameters (Polleri, paragraph [0031] “ Machine-learning configuration and interaction with the model composition engine 132 allows for selection of various library components 168 (e.g., pipelines 136 or workflows, micro services routines 140, software modules 144, and infrastructure modules 148) to define implementation of the logic of training and inference to build machine-learning applications 112. Different parameters, variables, scaling, settings, etc. for the library components 168 can be specified or determined by the model composition engine 132. The complexity conventionally required to create the machine-learning applications 112 can be performed largely automatically with the model composition engine 132.” Sharma paragraph [0268] “ A user can select a workflow from the list 1104 to see the details of the various components making up the workflow, which may include one or more machine learning models. A user can edit the parameters associated with a machine learning model that is already part of an existing workflow and also can identify and add a new machine learning model to an existing workflow. For example, a user can enter certain machine learning model information and parameters 1106, such as model type (e.g., regression), algorithm type (e.g., decision tree), and algorithm parameters, using the user interface. These parameters are stored in the software platform.”).
Regarding claim 4, Polleri in view of Sharma teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Polleri in view of Sharma further teaches:
providing lifecycle services of caching (Sharma paragraph [0114] “The time-series database (TSDB) is a software system that is optimized for handling time series data comprising arrays of numbers indexed by time (e.g., a date-time or a date-time range). The time-series database typically comprises a rolling or circular buffer or queue, in which as new information is added to the database, the oldest information is removed. This arrangement is beneficial for use in an edge platform environment because it makes efficient use of the limited storage capacity typically available in such an environment.” – the TSDB uses a rolling or circular buffer, where older data is discarded as new data arrives – this is the functional equivalent to a cache that keeps only the most recent and relevant data in fast local storage.), versioning (Sharma paragraph [0231]-[0233] “Iterative Closed-Loop Updating of Edge-Based ML Models. Once an edge-converted ML model is deployed to the edge platform and begins operating on live sensor data, it may be desirable to periodically evaluate the accuracy of the predictions, inferences, and other outputs generated by the model and iteratively update the model as necessary… FIGS. 8 and 9, a closed-loop arrangement between the edge platform 406, 609 and the cloud platform 412 provides for periodic evaluation and iterative updating of ML models on the edge platform.” – versioning at a local site.), and governance at the local site (Sharma, paragraph [0118] “The features and functionality of the edge platform are accessible via one or more applications 633” [0124] “As mentioned above, many different edge applications are possible, including apps that provide integrated administration and management 640 of one or more edge platforms, which may include monitoring or storing of data in the cloud or at a private data center 644. In addition, applications of the edge infrastructure can provide such important functionalities as real-time feedback and automated systems control to some of the toughest and most remote industrial environments. A number of specific applications are described in FogHorn's U.S. Pat. No. 10,007,513 at col. 13, line 55 to col. 15, line 13.” – teaches governance at a local site by describing integrated administration, management of edge platforms, including local control of data and system operation.)
Regarding claim 5, Polleri in view of Sharma teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Polleri in view of Sharma further teaches:
wherein a machine learning pipeline is used to perform the training of the AI model (Polleri [0031] “ Machine-learning configuration and interaction with the model composition engine 132 allows for selection of various library components 168 (e.g., pipelines 136 or workflows, micro services routines 140, software modules 144, and infrastructure modules 148) to define implementation of the logic of training and inference to build machine-learning applications 112.” [0034] “The machine-learning platform 100 can generate highly customizable applications. The library components 168 contain a set of predefined, off-the-shelf workflows or pipelines 136, which the application developer can incorporate into a new machine-learning application 112.” [0075] “At 380, process 300 can include training the machine-learning application using the selected machine-learning model template and the transformed dataset.” – teaches using a machine-learning pipeline 136 as part of the library components that implement and support model training.).
Regarding claim 6, Polleri in view of Sharma teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Polleri in view of Sharma further teaches:
receiving feedback on the AI model; and updating the AI model template and updating the AI model based on the feedback (Polleri, [0039] “The monitoring engine 156 can receive the results of the model execution engine 108 and compare the results with the performance characteristics (e.g., KPI/QoS metrics 160). The monitoring engine 156 can use ground truth data to test the machine-learning application 112 to ensure the model can perform as intended. The monitoring engine 156 can provide feedback to the model composition engine 132. The feedback can include adjustments to one or more variables or selected machine-learning model used in the machine-learning model 112.”).
Regarding claim 8, Polleri in view of Sharma teaches the following limitations:
A computer program product, the computer program product comprising a computer readable storage medium having program code embodied therewith, the program code executable by at least one processor to perform operations for (Polleri, paragraph [0141] “ In various embodiments, processing unit 604 can execute a variety of programs in response to program code and can maintain multiple concurrently executing programs or processes. At any given time, some or all of the program code to be executed can be resident in processing unit 604 and/or in storage subsystem 618…Computer system 600 may additionally include a processing acceleration unit 606, which can include a digital signal processor (DSP), a special-purpose processor, and/or the like.” Sharma paragraph [0055] “ A computer-implemented or computer-executable version or computer program product incorporating the invention or aspects thereof may be embodied using, stored on, or associated with computer-readable medium. A computer-readable medium may include any medium that participates in providing instructions to one or more processors for execution.”):
receiving selection of an Artificial Intelligence (Al) model template at a local site, wherein the AI model template is created at a remote site (Polleri, paragraph [0012] “According to some implementations, a server system may…perform operations to: receive an input, wherein the input identifies a problem to be solved using the machine-learning application…to select a machine-learning model template from a plurality of templates based at least in part on the input, wherein the machine-learning model template includes metadata, the metadata specifies data expectations and available data formats.” Paragraph [0035] “A model composition engine 132… can receive inputs from a user 116 through an interface 104…The model composition engine 132 can interface with library components 168 to identify various pipelines 136, micro service routines 140, software modules 144, and infrastructure models 148 that can be used in the creation of the machine-learning model 112.” – the model template is selected and created at the remote site (server) based on user input (local site). and is packaged in a transportable container (Polleri, paragraph [0036] “ The model composition engine 132 can output one or more machine-learning applications 112. The machine-learning applications 112 can be stored locally on a server or in a cloud-based network. The model composition engine 132 can output the machine-learning application 112 as executable code that be run on various infrastructure 128 through the infrastructure interfaces 124.” Sharma paragraph [0105] “Containers were developed as a part of the popular Linux open-source operating system…Docker containers wrap up a piece of software in a complete file system that contains everything the software needs to run: code, runtime, system tools, and system libraries—anything that can be installed on a server. This ensures the software will always run the same, regardless of the environment in which it is running. Thus, by incorporating a container technology such as Docker in the SDK, applications developed using the SDK can be deployed and executed not only on the edge platform for which they were developed, but also on other edge platform implementations, as well as in the cloud.”)
retrieving the AI model template in the transportable container from the remote site (Polleri [0065] “The selected machine-learning model can be selected for the machine-learning application. The selected machine-learning model can be downloaded and stored locally. In various embodiments, the system can inform the user of the specific model name or identifier of the selected model.” [0098] “Server 412 may be communicatively coupled with remote client computing devices 402, 404, 406, and 408 via network 410.” Sharma [0106] “In the example embodiment, containers for applications are created, deployed, and managed intelligently based on factors such as the types of edge environments and devices involved. This intelligence may be implemented in the form of a software apparatus comprising a number of components including centralized administration, deployment topology templates, container mobility management, zero-touch edge deployment, container monitoring, and responsive migration.” – Polleri teaches retrieving the AI model template from a remote site by disclosing that the selected machine-learning model is downloaded and stored locally. Polleri further describes a distributed network architecture in which server systems and client devices are communicatively coupled via a network. Under BRI, downloading selected model from a remote server system to a local storage reasonably suggests retrieving the AI model template from a remote site. Sharma further teaches that machine-learning applications are packaged and deployed using transportable containers, including “container mobility management” and “zero-touch edge deployment.” These teachings support transporting and deploying containerized model components across distributed environments, which reasonably suggest retrieving the AI model template in a transportable container from a remote site. Accordingly, the combination of Polleri and Sharma teaches or suggests retrieving the AI model template in the transportable container from the remote site.);
orchestrating a lifecycle of an AI model by: instantiating the AI model from the AI model template at the local site(Polleri, paragraph [0036] “The model composition engine 132 can output the machine-learning application 112 as executable code that be run on various infrastructure 128 through the infrastructure interfaces 124.” Paragraph [0037] “The model execution engine 108 can execute the machine-learning application 112 on infrastructure 128 using one or more the infrastructure interfaces 124.” [0044] “The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” [0065] “ The selected machine-learning model can be downloaded and stored locally.” – Polleri teaches instantiating the AI model from the AI model template at a local site. Polleri discloses that the machine-learning application is output as executable code and executed on infrastructure 128 through infrastructure inferences. Polleri further teaches that the data storage location may be local and electrically connected to the processing circuitry used to “generate, test, and execute the application,” and that the selected model can be downloaded and stored locally. Under the broadest reasonable interpretation, executing a locally stored machine-learning application on local infrastructure using locally connected processing circuitry reasonably suggests instantiating an AI model from the AI model template at the local site.);
retrieving data from one or more local data sources at the local site (Polleri, paragraph [0044] “The machine-learning platform 100 can include one or more data storage locations 170. The user can identify the one or more data storage locations 170. The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” [0046] “The model execution engine 108 can use hosted input data 164 to execute and test the machine-learning application 112. The hosted input data 164 can include a portion of the data stored at the data storage 170.” – Polleri teaches retrieving data from one or more local data sources at the local site by teaching that the machine-learning platform includes data storage locations that “can be local” and electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application. Polleri further teaches that the model execution engine uses hosted input data to execute and test the machine-learning application, where the hosted input data includes data stored at the data storage location 170. Under BRI, accessing and using data stored at locally connected data storage locations to execute and test the machine-learning application reasonably suggests retrieving data from one or more local data sources at the local site.)
training the AI model using the data at the local site (Polleri, [0044] “ The machine-learning platform 100 can include one or more data storage locations 170…The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” Paragraph [0054] “At 206, the model composition engine 132 can match one or more of the key features of the selected machine-learning model with a customer data schema of the customer data 208. In this way the customer data 208 can be used with the selected machine-learning model.” Paragraph [0056] “At 214, the model composition engine 132 can capture metrics for the customer data. The model can be trained and validated using known values.” – Polleri teaches using customer data with the selected machine-learning model and further teaches that the model can be trained and validated using known values. Polleri also teaches that the customer data maybe be stored in a local data storage location that is electrically connected to the processing circuitry used to generate, test, and execute the application. Under BRI, training the model using customer data that is stored locally and used by the model composition engine really corresponds to training the AI using the data at the local site. Accordingly, Polleri teaches or suggests training the AI model using the data at the local site.);
deploying the AI model as a service to generate one or more predictions and corresponding explanations at the local site (Polleri paragraph [0094] “ In some implementations, process 300 includes deploying the machine-learning application to a client infrastructure. Service deployment can start with a standard configuration as defined in the machine model template.” Paragraph [0099] “ In various embodiments, server 412 may be adapted to run one or more services or software applications provided by one or more of the components of the system. In some embodiments, these services may be offered as web-based or cloud services or under a Software as a Service (SaaS) model to the users of client computing devices 402, 404, 406, and/or 408.” Sharma, paragraph [0183] “Various predictions, inferences, analytical results, and other intelligence information, e.g., business and operational insights produced by applications, analytics expressions, and machine learning models executing on the edge platform” – Polleri teaches deploying the AI model as a service by describing deployment of the machine-learning application to a client infrastructure and operation of the application as a service available to client devices. Sharma teaches that machine-learning models executing on the edge platform generating predictions together with associated analytical results, intelligence information, and business and operational insights. Under BRI, generating predictions together with such intelligence information reasonably suggests providing accompanying contextual or interpretive information corresponding to the prediction output. Because Sharma locates these outputs on the edge platform, execution on the edge platform corresponds to execution at a local site. Accordingly, the combination of Polleri and Sharma teaches or suggests deploying the AI model as a service to generate predictions and corresponding explanatory information at a local site.);
monitoring the AI model for drift at the local site (Polleri, paragraph [0032] “A monitoring engine 156 can monitor operation of the machine-learning applications 112 according to the key performance indicators (KPI)/Quality of Service (QoS) metrics 160 to assure the machine-learning application 112 is performing according to requirements.” Sharma paragraph [0023] “A model update cycle may be initiated by a trigger, which may comprise a manual trigger, a time-based trigger, or a trigger derived from evaluating inferences generated by the model at the edge. Analytics expressions implementing selected logic, math, statistical or other functions may be applied to a stream of inferences generated by the model on the edge platform. The analytics expressions define what constitutes an unacceptable level of drift or degradation of model accuracy and track the model output to determine if the accuracy has degraded beyond an acceptable limit.” Polleri teaches monitoring operations of a machine-learning application using KPI and QoS metrics to ensure that the application performs according to the requirements. Sharma further teaches evaluating inferences generated at the edge and applying analytics expressions to determine whether the model accuracy has experienced drift or degradation beyond an acceptable limit. Under BRI, evaluating model outputs and determining whether model accuracy has degraded beyond a threshold reasonably corresponds to monitoring the AI model for drift. Because Sharma locates the inferencing, analytics evaluation, and drift determination on the edge platform, executing and monitoring on the edge platform reasonably correspond to monitoring the AI model for drift at the local site. Accordingly, the combination of Polleri and Sharma teaches or suggests monitoring the AI model for drift at the local site.); and
in response to identifying the drift, re-training the AI model at the local site (Sharma, paragraph [0023] “A model update cycle may be initiated by a trigger…The analytics expressions define what constitutes an unacceptable level of drift or degradation of model accuracy …In response, the edge platform can automatically take action…” [0024] “In an implementation of another aspect of the invention, dynamic non-disruptive machine learning model update and replacement on the edge computing platform is provided…a machine learning model on the edge computing platform may be updated with a modified machine learning model…” Polleri, paragraph [0094] “After a few cycles of training and serving, data and performance metrics can be used to automatically adjust the deployment configuration…” – Sharma teaches initiating a model update cycle in response to determining that model accuracy has experienced drift or degradation beyond and acceptable limit. Sharma further teaches updating and replacing the machine-learning model without interrupting ongoing inference operations. Under BRI, updating the model on the edge platform in response to identified drift reasonably corresponds to re-training the AI model at the local site. Polleri additionally teaches iterative cycles of training and serving in which performance metrics drive automatic adjustments to model behavior, reinforcing that model updates may be performed based on observed performance characteristics. Because Sharma expressly locates the inferencing, drift evaluation, and model update operations on the edge computing platform, the edge platform reasonably corresponds to the claimed local site. Accordingly, the combination of Sharma and Polleri teaches or suggests re-training the AI model at the local site in response to identifying drift.)
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having a combination of Polleri and Sharma before them, to incorporate the use of an advanced software container for ensuring transportability, as detailed by Sharma, into the model deployment system described by Polleri. One would have been motivated to make such a combination because Polleri details outputting the machine learning application as executable code that can be run on various infrastructure, and Sharma teaches that this self-contained packaging wraps up the software and its dependencies to ensure it will always run the same, regardless of the environment in which it is running. This capability is necessary to produce a reliable, universally deployable, and easily transportable container that overcomes configuration barriers at local sites.
Regarding claim 9, Polleri in view of Sharma teaches all the elements of claim 8, therefore is rejected for the same reasons as those presented for claim 8. The claim recites similar limitations corresponding to claim 2. Therefore, the claim is rejected for similar reasons as claim 2 using similar teachings and rationale.
Regarding claim 10, Polleri in view of Sharma teaches all the elements of claim 8, therefore is rejected for the same reasons as those presented for claim 8. The claim recites similar limitations corresponding to claim 3. Therefore, the claim is rejected for similar reasons as claim 3 using similar teachings and rationale.
Regarding claim 11, Polleri in view of Sharma teaches all the elements of claim 8, therefore is rejected for the same reasons as those presented for claim 8. The claim recites similar limitations corresponding to claim 4. Therefore, the claim is rejected for similar reasons as claim 4 using similar teachings and rationale.
Regarding claim 12, Polleri in view of Sharma teaches all the elements of claim 8, therefore is rejected for the same reasons as those presented for claim 8. The claim recites similar limitations corresponding to claim 5. Therefore, the claim is rejected for similar reasons as claim 5 using similar teachings and rationale.
Regarding claim 13, Polleri in view of Sharma teaches all the elements of claim 8, therefore is rejected for the same reasons as those presented for claim 8. The claim recites similar limitations corresponding to claim 6. Therefore, the claim is rejected for similar reasons as claim 6 using similar teachings and rationale.
Regarding claim 15, Polleri teaches the following limitations:
A computer system, comprising: one or more processors, one or more computer-readable memories and one or more computer-readable, tangible storage devices; and program instructions, stored on at least one of the one or more computer-readable, tangible storage devices for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to perform operations comprising (Polleri, [0147] “Storage subsystem 618 may also provide a tangible computer-readable storage medium for storing the basic programming and data constructs that provide the functionality of some embodiments. Software (programs, code modules, instructions) that when executed by a processor provide the functionality described above may be stored in storage subsystem 618. These software modules or instructions may be executed by processing unit 604. Storage subsystem 618 may also provide a repository for storing data used in accordance with the present disclosure.”):
orchestrating a lifecycle of an AI model by: instantiating the AI model from the AI model template at the local site (Polleri, paragraph [0036] “The model composition engine 132 can output the machine-learning application 112 as executable code that be run on various infrastructure 128 through the infrastructure interfaces 124.” Paragraph [0037] “The model execution engine 108 can execute the machine-learning application 112 on infrastructure 128 using one or more the infrastructure interfaces 124.” [0044] “The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” [0065] “ The selected machine-learning model can be downloaded and stored locally.” – Polleri teaches instantiating the AI model from the AI model template at a local site. Polleri discloses that the machine-learning application is output as executable code and executed on infrastructure 128 through infrastructure inferences. Polleri further teaches that the data storage location may be local and electrically connected to the processing circuitry used to “generate, test, and execute the application,” and that the selected model can be downloaded and stored locally. Under the broadest reasonable interpretation, executing a locally stored machine-learning application on local infrastructure using locally connected processing circuitry reasonably suggests instantiating an AI model from the AI model template at the local site.);
retrieving data from one or more local data sources at the local site (Polleri, paragraph [0044] “The machine-learning platform 100 can include one or more data storage locations 170. The user can identify the one or more data storage locations 170. The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” [0046] “The model execution engine 108 can use hosted input data 164 to execute and test the machine-learning application 112. The hosted input data 164 can include a portion of the data stored at the data storage 170.” – Polleri teaches retrieving data from one or more local data sources at the local site by teaching that the machine-learning platform includes data storage locations that “can be local” and electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application. Polleri further teaches that the model execution engine uses hosted input data to execute and test the machine-learning application, where the hosted input data includes data stored at the data storage location 170. Under BRI, accessing and using data stored at locally connected data storage locations to execute and test the machine-learning application reasonably suggests retrieving data from one or more local data sources at the local site.)
training the AI model using the data at the local site (Polleri, [0044] “ The machine-learning platform 100 can include one or more data storage locations 170…The data storage location 170 can be local (e.g., in a storage device electrically connected to the processing circuitry and interfaces used to generate, test, and execute the application).” Paragraph [0054] “At 206, the model composition engine 132 can match one or more of the key features of the selected machine-learning model with a customer data schema of the customer data 208. In this way the customer data 208 can be used with the selected machine-learning model.” Paragraph [0056] “At 214, the model composition engine 132 can capture metrics for the customer data. The model can be trained and validated using known values.” – Polleri teaches using customer data with the selected machine-learning model and further teaches that the model can be trained and validated using known values. Polleri also teaches that the customer data maybe be stored in a local data storage location that is electrically connected to the processing circuitry used to generate, test, and execute the application. Under BRI, training the model using customer data that is stored locally and used by the model composition engine really corresponds to training the AI using the data at the local site. Accordingly, Polleri teaches or suggests training the AI model using the data at the local site.);
However, Polleri does not teach, but, Polleri in view of Sharma teaches the following limitations:
deploying the AI model as a service to generate one or more predictions and corresponding explanations at the local site (Polleri paragraph [0094] “ In some implementations, process 300 includes deploying the machine-learning application to a client infrastructure. Service deployment can start with a standard configuration as defined in the machine model template.” Paragraph [0099] “ In various embodiments, server 412 may be adapted to run one or more services or software applications provided by one or more of the components of the system. In some embodiments, these services may be offered as web-based or cloud services or under a Software as a Service (SaaS) model to the users of client computing devices 402, 404, 406, and/or 408.” Sharma, paragraph [0183] “Various predictions, inferences, analytical results, and other intelligence information, e.g., business and operational insights produced by applications, analytics expressions, and machine learning models executing on the edge platform” – Polleri teaches deploying the AI model as a service by describing deployment of the machine-learning application to a client infrastructure and operation of the application as a service available to client devices. Sharma teaches that machine-learning models executing on the edge platform generating predictions together with associated analytical results, intelligence information, and business and operational insights. Under BRI, generating predictions together with such intelligence information reasonably suggests providing accompanying contextual or interpretive information corresponding to the prediction output. Because Sharma locates these outputs on the edge platform, execution on the edge platform corresponds to execution at a local site. Accordingly, the combination of Polleri and Sharma teaches or suggests deploying the AI model as a service to generate predictions and corresponding explanatory information at a local site.);
receiving selection of an Artificial Intelligence (Al) model template at a local site, wherein the AI model template is created at a remote site (Polleri, paragraph [0012] “According to some implementations, a server system may…perform operations to: receive an input, wherein the input identifies a problem to be solved using the machine-learning application…to select a machine-learning model template from a plurality of templates based at least in part on the input, wherein the machine-learning model template includes metadata, the metadata specifies data expectations and available data formats.” Paragraph [0035] “A model composition engine 132… can receive inputs from a user 116 through an interface 104…The model composition engine 132 can interface with library components 168 to identify various pipelines 136, micro service routines 140, software modules 144, and infrastructure models 148 that can be used in the creation of the machine-learning model 112.” – the model template is selected and created at the remote site (server) based on user input (local site). and is packaged in a transportable container (Polleri, paragraph [0036] “ The model composition engine 132 can output one or more machine-learning applications 112. The machine-learning applications 112 can be stored locally on a server or in a cloud-based network. The model composition engine 132 can output the machine-learning application 112 as executable code that be run on various infrastructure 128 through the infrastructure interfaces 124.” Sharma paragraph [0105] “Containers were developed as a part of the popular Linux open-source operating system…Docker containers wrap up a piece of software in a complete file system that contains everything the software needs to run: code, runtime, system tools, and system libraries—anything that can be installed on a server. This ensures the software will always run the same, regardless of the environment in which it is running. Thus, by incorporating a container technology such as Docker in the SDK, applications developed using the SDK can be deployed and executed not only on the edge platform for which they were developed, but also on other edge platform implementations, as well as in the cloud.”
retrieving the AI model template in the transportable container from the remote site(Polleri [0065] “The selected machine-learning model can be selected for the machine-learning application. The selected machine-learning model can be downloaded and stored locally. In various embodiments, the system can inform the user of the specific model name or identifier of the selected model.” [0098] “Server 412 may be communicatively coupled with remote client computing devices 402, 404, 406, and 408 via network 410.” Sharma [0106] “In the example embodiment, containers for applications are created, deployed, and managed intelligently based on factors such as the types of edge environments and devices involved. This intelligence may be implemented in the form of a software apparatus comprising a number of components including centralized administration, deployment topology templates, container mobility management, zero-touch edge deployment, container monitoring, and responsive migration.” – Polleri teaches retrieving the AI model template from a remote site by disclosing that the selected machine-learning model is downloaded and stored locally. Polleri further describes a distributed network architecture in which server systems and client devices are communicatively coupled via a network. Under BRI, downloading selected model from a remote server system to a local storage reasonably suggests retrieving the AI model template from a remote site. Sharma further teaches that machine-learning applications are packaged and deployed using transportable containers, including “container mobility management” and “zero-touch edge deployment.” These teachings support transporting and deploying containerized model components across distributed environments, which reasonably suggest retrieving the AI model template in a transportable container from a remote site. Accordingly, the combination of Polleri and Sharma teaches or suggests retrieving the AI model template in the transportable container from the remote site.);
monitoring the AI model for drift at the local site (Polleri, paragraph [0032] “A monitoring engine 156 can monitor operation of the machine-learning applications 112 according to the key performance indicators (KPI)/Quality of Service (QoS) metrics 160 to assure the machine-learning application 112 is performing according to requirements.” Sharma paragraph [0023] “A model update cycle may be initiated by a trigger, which may comprise a manual trigger, a time-based trigger, or a trigger derived from evaluating inferences generated by the model at the edge. Analytics expressions implementing selected logic, math, statistical or other functions may be applied to a stream of inferences generated by the model on the edge platform. The analytics expressions define what constitutes an unacceptable level of drift or degradation of model accuracy and track the model output to determine if the accuracy has degraded beyond an acceptable limit.” Polleri teaches monitoring operations of a machine-learning application using KPI and QoS metrics to ensure that the application performs according to the requirements. Sharma further teaches evaluating inferences generated at the edge and applying analytics expressions to determine whether the model accuracy has experienced drift or degradation beyond an acceptable limit. Under BRI, evaluating model outputs and determining whether model accuracy has degraded beyond a threshold reasonably corresponds to monitoring the AI model for drift. Because Sharma locates the inferencing, analytics evaluation, and drift determination on the edge platform, executing and monitoring on the edge platform reasonably correspond to monitoring the AI model for drift at the local site. Accordingly, the combination of Polleri and Sharma teaches or suggests monitoring the AI model for drift at the local site.); and
in response to identifying the drift, re-training the AI model at the local site (Sharma, paragraph [0023] “A model update cycle may be initiated by a trigger…The analytics expressions define what constitutes an unacceptable level of drift or degradation of model accuracy …In response, the edge platform can automatically take action…” [0024] “In an implementation of another aspect of the invention, dynamic non-disruptive machine learning model update and replacement on the edge computing platform is provided…a machine learning model on the edge computing platform may be updated with a modified machine learning model…” Polleri, paragraph [0094] “After a few cycles of training and serving, data and performance metrics can be used to automatically adjust the deployment configuration…” – Sharma teaches initiating a model update cycle in response to determining that model accuracy has experienced drift or degradation beyond and acceptable limit. Sharma further teaches updating and replacing the machine-learning model without interrupting ongoing inference operations. Under BRI, updating the model on the edge platform in response to identified drift reasonably corresponds to re-training the AI model at the local site. Polleri additionally teaches iterative cycles of training and serving in which performance metrics drive automatic adjustments to model behavior, reinforcing that model updates may be performed based on observed performance characteristics. Because Sharma expressly locates the inferencing, drift evaluation, and model update operations on the edge computing platform, the edge platform reasonably corresponds to the claimed local site. Accordingly, the combination of Sharma and Polleri teaches or suggests re-training the AI model at the local site in response to identifying drift.)
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having a combination of Polleri and Sharma before them, to incorporate the use of an advanced software container for ensuring transportability, as detailed by Sharma, into the model deployment system described by Polleri. One would have been motivated to make such a combination because Polleri details outputting the machine learning application as executable code that can be run on various infrastructure, and Sharma teaches that this self-contained packaging wraps up the software and its dependencies to ensure it will always run the same, regardless of the environment in which it is running. This capability is necessary to produce a reliable, universally deployable, and easily transportable container that overcomes configuration barriers at local sites.
Regarding claim 16, Polleri in view of Sharma teaches all the elements of claim 15, therefore is rejected for the same reasons as those presented for claim 15. The claim recites similar limitations corresponding to claim 2, therefore, the claim is rejected for similar reasons as claim 2 using similar teachings and rationale.
Regarding claim 17, Polleri in view of Sharma teaches all the elements of claim 15, therefore is rejected for the same reasons as those presented for claim 15. The claim recites similar limitations corresponding to claim 3, therefore, the claim is rejected for similar reasons as claim 3 using similar teachings and rationale.
Regarding claim 18, Polleri in view of Sharma teaches all the elements of claim 15, therefore is rejected for the same reasons as those presented for claim 15. The claim recites similar limitations corresponding to claim 4, therefore, the claim is rejected for similar reasons as claim 4 using similar teachings and rationale.
Regarding claim 19, Polleri in view of Sharma teaches all the elements of claim 15, therefore is rejected for the same reasons as those presented for claim 15. The claim recites similar limitations corresponding to claim 5, therefore, the claim is rejected for similar reasons as claim 5 using similar teachings and rationale.
Regarding claim 20, Polleri in view of Sharma teaches all the elements of claim 15, therefore is rejected for the same reasons as those presented for claim 15. The claim recites similar limitations corresponding to claim 6, therefore, the claim is rejected for similar reasons as claim 6 using similar teachings and rationale.
Claims 7 and 14 are rejected under the 35 U.S.C. 103 as being unpatentable over Polleri et al., (Pub No.: US 20210081720 A1 (Filed: 2020)) in view of Sharma et al., (Pub. No.: US 20200327371 A1 (Filed: 2019)) further in view of Zhao et al., (NPL: “Packaging and Sharing Machine Learning Models via the Acumos AI Open Platform” (Published: 2018)).
Regarding claim 7, Polleri in view of Sharma teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. However, Polleri in view of Sharma does not teach, but Polleri in view of Sharma further in view of Zhao teaches the following limitation:
wherein the AI model template is a formalized description of the AI model that allows transportability of the model, and wherein code, AI model information, an AI model manifest, pipeline metadata, and binary execution images are packaged in the AI model template (Polleri, paragraph [0049] “Each of the machine-learning models can include metadata that identifies the type of machine-learning model and indicate the type of problems the machine-learning model is intended to solve.” [0036] “The model composition engine 132 can output the machine-learning application 112 as executable code that be run on various infrastructure 128 through the infrastructure interfaces 124.” Zhao, [page 843] “the contributor is required to describe the model’s metadata when publishing, such as the function description, input and output format and model category.” & “The Acumos platform will pack the uploaded model as a microservice in a Docker image…” & “The Acumos platform packs the model into a Docker image which can be deployed to an appropriate run-time environment.” & “Acumos packages solutions into Docker images which can then be deployed into any Docker environment” – Polleri teaches executable model code and model-related metadata associated with the machine-learning model, which reasonably correspond to the claimed code and AI model information. Zhao further teaches describing model metadata including function descriptions, input/output formats, and model categories, which reasonably correspond to an AI model manifest or descriptive information associated with the model. Zhao additionally teaches packaging uploaded models and solutions into Docker images deployable across different runtime environments, which reasonably correspond to binary execution images packed in a transportable AI model template. Under BRI, packaging executable model code, model information, descriptive metadata, and Docker runtime artifacts into deployable Docker images reasonably suggests a formalized description of the AI model that allows transportability of the model across execution environments. Accordingly, the combination of Polleri and Zhao teaches or suggest the limitation.).
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having a combination of Polleri, Sharma, and Zhao before them, to further clarify that the docker container of Sharma is a Docker binary execution image used in the art for packaging model code, metadata, and run-time artifacts into a portable, transportable format.
Regarding claim 14, Polleri in view of Sharma teaches all the elements of claim 8, therefore is rejected for the same reasons as those presented for claim 8. However, Polleri in view of Sharma does not teach, but Polleri in view of Sharma further in view of Zhao teaches the following limitation:
wherein the AI model template is a formalized description of the AI model that allows transportability of the model, and wherein code, AI model information, an AI model manifest, pipeline metadata, and binary execution images are packaged in the AI model template (Polleri, paragraph [0049] “Each of the machine-learning models can include metadata that identifies the type of machine-learning model and indicate the type of problems the machine-learning model is intended to solve.” [0036] “The model composition engine 132 can output the machine-learning application 112 as executable code that be run on various infrastructure 128 through the infrastructure interfaces 124.” Zhao, [page 843] “the contributor is required to describe the model’s metadata when publishing, such as the function description, input and output format and model category.” & “The Acumos platform will pack the uploaded model as a microservice in a Docker image…” & “The Acumos platform packs the model into a Docker image which can be deployed to an appropriate run-time environment.” & “Acumos packages solutions into Docker images which can then be deployed into any Docker environment” – Polleri teaches executable model code and model-related metadata associated with the machine-learning model, which reasonably correspond to the claimed code and AI model information. Zhao further teaches describing model metadata including function descriptions, input/output formats, and model categories, which reasonably correspond to an AI model manifest or descriptive information associated with the model. Zhao additionally teaches packaging uploaded models and solutions into Docker images deployable across different runtime environments, which reasonably correspond to binary execution images packed in a transportable AI model template. Under BRI, packaging executable model code, model information, descriptive metadata, and Docker runtime artifacts into deployable Docker images reasonably suggests a formalized description of the AI model that allows transportability of the model across execution environments. Accordingly, the combination of Polleri and Zhao teaches or suggest the limitation.).
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having a combination of Polleri, Sharma, and Zhao before them, to further clarify that the docker container of Sharma is a Docker binary execution image used in the art for packaging model code, metadata, and run-time artifacts into a portable, transportable format.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 Daravanh Phakousonh whose telephone number is (571)272-6324. The examiner can normally be reached Mon - Thurs 7 AM - 5 PM, Every other Friday 7 AM - 4PM.
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, Li B Zhen can be reached at 571-272-3768. 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.
/Daravanh Phakousonh/Examiner, Art Unit 2121
/Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights