Method for decentralized accessioning for distributed machine learning and other applications

Patent metadata

Field	Value
Publication number	US-12561223-B2
Application number	US-202017002444-A
Country	US
Kind code	B2
Filing date	Aug 25, 2020
Priority date	Aug 25, 2020
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

A method for injecting metadata into an existing artifact is described. The method generates metadata related to an existing artifact having a predetermined structure and encodes the metadata in accordance with the predetermined structure. The encoded metadata is embedded within the existing artifact in accordance with the predetermined structure and is delineated within the predetermined structure as one or more individual records. The artifact, including embedded metadata, is stored within a storage entity and is accessible to processes related to the artifact. Additional records may be generated and embedded over time, thus creating a timeline if event related to the artifact.

First claim

The invention claimed is:

1 . A computer-implemented method for injecting metadata into an existing machine learning process artifact comprising:
generating metadata related to an existing machine learning process artifact during one or more processes related to the existing machine learning process, wherein the existing machine learning process artifact is embodied in a predetermined structure and further wherein the metadata is generated at multiple distinct times; encoding the metadata at each distinct time in accordance with the predetermined structure; and embedding the metadata at each distinct time within the existing machine learning process artifact in accordance with the predetermined structure, wherein the embedded metadata at each distinct time is delineated within the predetermined structure as a separate individual record; storing the existing machine learning process artifact with embedded separate individual records in a storage entity after each distinct time as a distinct version.

2 . The computer-implemented method according to claim 1 , wherein the processes related to the existing machine learning process artifact are selected from a group consisting of creating, duplicating, training, serving, deploying, producing, storing and using the existing machine learning process artifact.

3 . The computer-implemented method according to claim 1 , wherein the separate individual records containing the metadata are injected into the existing machine learning process artifact over time to create an audit trail.

4 . The computer-implemented method according to claim 3 , wherein the separate individual records are injected sequentially to the existing machine learning process artifact over time.

5 . The computer-implemented method according to claim 1 , wherein the one or more processes are performed by one or more processing entities selected from a group consisting of instrumented and un-instrumented processing entities.

6 . The computer-implemented method according to claim 5 , wherein metadata generated by a process performed by an instrumented processing entity is encoded and embedded within the existing machine learning process artifact by the instrumented processing entity.

7 . The computer-implemented method according to claim 5 , wherein metadata generated by a process performed by an un-instrumented processing entity is collected by a supervising entity wherein the supervising entity encodes and embeds the metadata within the existing machine learning process artifact on behalf of the un-instrumented processing entity.

8 . The computer-implemented method according to claim 1 , wherein the existing machine learning process artifact is selected from a group consisting of data, code and model artifacts.

9 . A computer-implemented method for injecting metadata generated during one or more processes related to a model into an existing machine learning process artifact comprising:
generating first metadata related to an existing machine learning process artifact during a first process, wherein the existing machine learning process artifact is embodied in a predetermined structure; encoding the first metadata in accordance with the predetermined structure; and embedding the first metadata within the existing machine learning process artifact in accordance with the predetermined structure, wherein the embedded first metadata is delineated within the predetermined structure as a first individual record; storing the existing machine learning process artifact with embedded first individual record in a first storage entity as a first updated version of the existing machine learning process artifact; generating second metadata related to the existing machine learning process artifact during a second process; encoding the second metadata in accordance with the predetermined structure; embedding the second metadata within the existing machine learning process artifact in accordance with the predetermined structure, wherein the embedded second metadata is delineated within the predetermined structure as a second individual record; storing the existing machine learning process artifact with embedded first and second individual records in a second storage entity as a second updated version of the existing machine learning process artifact.

10 . The computer-implemented method according to claim 9 , wherein the first and second process are selected from a group consisting of duplicating, training, serving, deploying, producing, storing and using the existing machine learning process artifact.

11 . The computer-implemented method according to claim 9 , wherein the first and second individual records containing the first and second metadata are injected into the existing machine learning process artifact at different times, thereby creating an audit trail related to the existing machine learning process artifact.

12 . The computer-implemented method according to claim 9 , wherein the first and second processes are performed by one or more processing entities selected from a group consisting of instrumented and un-instrumented processing entities.

13 . The computer-implemented method according to claim 12 , wherein first and second metadata generated by processes performed by an instrumented processing entity is encoded and embedded within the existing machine learning process artifact by the instrumented processing entity.

14 . The computer-implemented method according to claim 13 , wherein first and second metadata generated by a process performed by an un-instrumented processing entity is collected by a supervising entity wherein the supervising entity encodes and embeds the metadata within the existing machine learning process artifact on behalf of the un-instrumented processing entity.

15 . The computer-implemented method according to claim 9 , wherein the first storage entity and the second storage entity are the same storage entity.

16 . The computer-implemented method according to claim 9 , wherein the first storage entity and the second storage entity are different storage entities.

17 . The computer-implemented method according to claim 9 , wherein the existing machine learning process artifact is selected from a group consisting of data, code and model artifacts.

18 . A computer-implemented method for automating access to and use of machine learning process artifact-related metadata from one or more sources by one or more processing frameworks, comprising:
generating metadata related to an existing machine learning process artifact, wherein the existing machine learning process artifact is embodied in a predetermined structure; encoding the metadata in accordance with the predetermined structure; embedding the metadata within the existing machine learning process artifact in accordance with the predetermined structure, wherein the embedded metadata is delineated within the predetermined structure as one or more individual records; storing the existing machine learning process artifact with embedded one or more individual records in a first source at a first time, wherein the first source includes a storage entity with an associated query engine for receiving queries related to the existing machine learning process artifact including the metadata embedded therein, and further wherein the storing results in creation of storage entity metadata associated with the existing machine learning process artifact wherein the storage entity metadata is encoded in accordance with the predetermined structure of the existing m

Assignees

• Cognizant Tech Solutions U S Corporation

Inventors

• Mutch Karl

Classifications

• G06N20/00

  Machine learning · CPC title

• G06N3/126

  Evolutionary algorithms, e.g. genetic algorithms or genetic programming · CPC title

• G06F11/3466 Primary

  Performance evaluation by tracing or monitoring · CPC title

• G06F16/164 Primary

  File meta data generation · CPC title