System for operationalizing high-level machine learning training enhancements from low-level primitives

Patent metadata

Field	Value
Publication number	US-2021011757-A1
Application number	US-201916510534-A
Country	US
Kind code	A1
Filing date	Jul 12, 2019
Priority date	Jul 12, 2019
Publication date	Jan 14, 2021
Grant date	—

Abstract

In an embodiment, a method for inspecting and transforming a machine learning model includes receiving a request that includes the machine learning model and a configuration object that provides an indication of a selected strategy. In the embodiment, the method includes creating a partially specified task graph that includes a first placeholder node for a future expanded task node. In the embodiment, the method includes performing a dynamic expansion and execution phase that includes, repeatedly (a) using a cognitive engine to evaluate whether to revise the partially specified task graph based at least in part on the selected strategy, and (b) using a processor-based execution engine to perform an action specified by the complete node. In an embodiment, the dynamic expansion and execution phase repeats until after the cognitive engine adds a consolidated results node.

First claim

What is claimed is:

1 . A computer implemented method for inspecting and transforming a machine learning model, the computer implemented method comprising:
receiving, by one or more processors, a request that includes the machine learning model and a configuration object, wherein the configuration object provides an indication of a selected strategy; creating, by one or more processors, a partially specified task graph based on the selected strategy, wherein the partially specified task graph includes a first placeholder node for a future expanded task node; and performing, by one or more processors, a dynamic expansion and execution phase that includes, repeatedly:
if at least one placeholder node is present in the partially specified task graph, using a cognitive engine to evaluate whether to revise the partially specified task graph based at least in part on the selected strategy; and
if the cognitive engine adds a complete node to the specified graph, using a processor-based execution engine to perform an action specified by the complete node;
wherein the dynamic expansion and execution phase repeats until after the cognitive engine adds a consolidated results node.

2 . The computer implemented method of claim 1 , further comprising:
receiving, via a graphical user interface (GUI) generated by one or more processors, user input data indicative of the selected strategy, the selected strategy corresponding to a displayed strategy option; and retrieving, from a computer-readable storage device by one or more processors, an algorithm library associated with the selected strategy.

3 . The computer implemented method of claim 2 , further comprising:
creating, by one or more processors, the configuration object based at least in part on the selected strategy and the algorithm library associated therewith.

4 . The computer implemented method of claim 3 , wherein the creating, by the one or more processors, of the partially specified task graph includes initializing the partially specified task graph to include a number of nodes, wherein the number of nodes is based on the algorithm library associated with the selected strategy.

5 . The computer implemented method of claim 1 , wherein the dynamic expansion and execution phase further comprises storing, by one or more processors in a computer-readable memory, historical data representative of a result of an action previously performed by the execution engine as specified by a previous node of the partially specified task graph.

6 . The computer implemented method of claim 5 , wherein the cognitive engine evaluates whether to revise the partially specified task graph based at least in part on the historical data and based on the configuration object.

7 . The computer implemented method of claim 1 , wherein the cognitive engine revises the partially specified task graph based on a revised hyperparameter.

8 . The computer implemented method of claim 1 , wherein the cognitive engine evaluates how to revise the partially specified task graph by querying a composition planning module to determine a next step in a composition plan for executing the request.

9 . The computer implemented method of claim 1 , wherein the cognitive engine revises the partially specified task graph according to a selected strategy for improving a robustness of the machine learning model.

10 . The computer implemented method of claim 1 , wherein the performing of the dynamic and execution phase includes a plurality of iterations of the using of the execution engine to perform respective actions, wherein the respective actions collectively modify the machine learning module and thereby repair model revisions previously made by malicious training data.

11 . A computer usable program product for inspecting and transforming a machine learning model, the computer usable program product comprising a computer-readable storage device, and program instructions stored on the storage device, the stored program instructions comprising:
program instructions to receive, by one or more processors, a request that includes the machine learning model and a configuration object, wherein the configuration object provides an indication of a selected strategy; program instructions to create, by one or more processors, a partially specified task graph based on the selected strategy, wherein the partially specified task graph includes a first placeholder node for a future expanded task node; and program instructions to perform, by one or more processors, a dynamic expansion and execution phase that includes, repeatedly:
if at least one placeholder node is present in the partially specified task graph, using a cognitive engine to evaluate whether to revise the partially specified task graph based at least in part on the selected strategy; and
if the cognitive engine adds a complete node to the specified graph, using a processor-based execution engine to perform an action specified by the complete node;
wherein the dynamic expansion and execution phase repeats until after the cognitive engine adds a consolidated results node.

12 . The computer usable program product of claim 11 , further comprising:
program instructions to receive, via a graphical user interface (GUI) generated by one or more processors, user input data indicative of the selected strategy, the selected strategy corresponding to a displayed strategy option; and program instructions to retrieve, from a computer-readable storage device by one or more processors, an algorithm library associated with the selected strategy.

13 . The computer implemented method of claim 12 , further comprising:
program instructions to create, by one or more processors, the configuration object based at least in part on the selected strategy and the algorithm library associated therewith.

14 . The computer implemented method of claim 13 , wherein the creating, by the one or more processors, of the partially specified task graph includes initializing the partially specified task graph to include a number of nodes, wherein the number of nodes is based on the algorithm library associated with the selected strategy.

15 . The computer implemented method of claim 11 , wherein the dynamic expansion and execution phase further comprises storing, by one or more processors in a computer-readable memory, historical data representative of a result of an action previously performed by the execution engine as specified by a previous node of the partially specified task graph.

16 . A computer system comprising a processor, a computer-readable memory, and a computer-readable storage device, and program instructions stored on the storage device for execution by the processor via the memory, the stored program instructions comprising:
program instructions to receive, by one or more processors, a request that includes the machine learning model and a configuration object, wherein the configuration object provides an indication of a selected strategy; program instructions to create, by one or more processors, a partially specified task graph based on the selected strategy, wherein the partially specified task graph includes a first placeholder node for a future expanded task node; and program instructions to perform, by one or more processors, a dynamic expansion and execution phase that includes, repeatedly:
if at least one placeholder node is present in the partially specified task graph, using a cognitive engine to evaluate whether to revise the partially specified task graph based at least in part on the selected strategy; and
if

Assignees

• IBM

Inventors

• Duesterwald Evelyn
• Murthi Anupama
• Vijaykeerthy Deepak
• Arya Vijay
• Venkataraman Ganesh

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• G06F9/4843 Primary

  by program, e.g. task dispatcher, supervisor, operating system · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06F18/24

  Classification techniques · CPC title