Providing custom machine-learning models

Patent metadata

Field	Value
Publication number	US-11475370-B2
Application number	US-201816205070-A
Country	US
Kind code	B2
Filing date	Nov 29, 2018
Priority date	Nov 29, 2018
Publication date	Oct 18, 2022
Grant date	Oct 18, 2022

Abstract

Providing custom machine learning models to client computer systems. Multiple machine learning models are accessed. Client-specific data for multiple client computer systems are also accessed. For each of at least some of the client computer systems, performing the following actions: First, using the corresponding client-specific data for the corresponding client computer system to determine which subset of the multiple machine learning models is applicable to the corresponding client computer system. The subset of the multiple machine learning models includes more than one of the multiple machine learning models. Then, aggregating the determined subset of the multiple machine learning models to generate an aggregated subset of machine learning models that is customized to the corresponding client computer system.

First claim

What is claimed is:

1. A computer system configured to provide custom machine learning models to a client computer system, said computer system comprising:
one or more processors; and
one or more computer-readable media having stored thereon computer-executable instructions that are executable by the one or more processors to cause the computer system to:
access a plurality of machine learning models;
access client-specific data for a client computer system;
use the client-specific data for the client computer system to determine which subset of the plurality of machine learning models is applicable to the client computer system, wherein:
the subset comprises multiple machine learning models selected from the plurality of machine learning models, and
determining which machine learning models are included in the subset is performed by: (i) determining attributes of source code associated with the client-specific data, (ii) filtering the plurality of machine learning models to identify particular models that are applicable to the client computer system based on the determined attributes, (iii) using machine learning to build an additional model that is applicable to the client computer system based on the determined attributes, and (iv) including the built model and said particular models in the subset; and

aggregate the determined subset to generate an aggregated subset of machine learning models that is customized to the client computer system.

2. The computer system of claim 1 , wherein:
the client computer system is configured to operate in at least one programming environment;
the aggregating of the determined subset comprises:
aggregating the determined subset at a machine learning service provider system; and
sending the aggregated subset to the client computer system.

3. The computer system of claim 2 , wherein the client computer system is a first client computer system, wherein for a second client computer system, the aggregating of the determined subset comprises:
sending the determined subset to the second client computer system; and
aggregating the determined subset on the second client computer system.

4. The computer system of claim 1 , wherein the client computer system is configured to operate in at least one programming environment;
wherein the aggregating of the determined subset comprises:
aggregating the determined subset at a machine learning service provider system;
accessing the client computer system;
operating the aggregated subset at the machine learning service provider system; and
providing a result of the operation of the aggregated subset to the client computer system.

5. The computer system of claim 1 , wherein aggregating the determined subset comprises:
sending the determined subset to the client computer system; and
aggregating the determined subset on the client computer system.

6. The computer system of claim 1 , wherein using the client-specific data for the client computer system to determine which subset of the plurality of machine learning models is applicable to the client computer system comprises:
for each of at least some machine learning models in the plurality of machine learning models:
determining whether the client computer system has permission to access said each machine learning model; and
if the client computer system has the permission, including the permitted corresponding machine learning model in the subset.

7. The computer system of claim 6 , wherein whether the client computer system has permission to access said each machine learning model is determined by a second client computer system.

8. The computer system of claim 1 , wherein using the client-specific data for the client computer system to determine which subset of the plurality of machine learning models is applicable to the client computer system comprises:
determining an application type of an application that the client computer system is working on; and
filtering the plurality of machine learning models based on the determined application type to determine which subset of the plurality of machine learning models is applicable to a datatype of the client-specific data.

9. The computer system of claim 8 , wherein the client-specific data is selected from a group consisting of programming language, application type, compiler, package reference, component reference, software developer's kit (SDK) references, code quality metrics, whether the source code is test code or application code, and version of programming frameworks.

10. The computer system of claim 1 , wherein using the client-specific data for the client computer system to determine which subset of the plurality of machine learning models is applicable to the client computer system comprises:
determining a datatype of the client-specific data; and
filtering the plurality of machine learning models based on the determined datatype of the client-specific data to determine which subset of the plurality of machine learning models is applicable to the datatype of the client-specific data.

11. The computer system of claim 10 , wherein the determined datatype is computer programming source code.

12. The computer system of claim 11 , wherein using the client-specific data for the client computer system to determine which subset of the plurality of machine learning models is applicable to the client computer system comprises:
compiling the computer programming source code;
building a semantic model based on the compiled computer programming code;
analyzing the semantic model to obtain a set of data features; and
determining which subset of the plurality of machine learning models is applicable to the client computer system based on the set of data features generated.

13. The computer system of claim 12 , wherein execution of the instructions further causes the computer system to:
apply the aggregated subset of machine learning models to the computer programming source code.

14. The computer system of claim 12 , wherein execution of the instructions further causes the computer system to:
in response to the application of the aggregated subset of machine learning models, predicting a programming element for the client computer system.

15. The computer system of claim 14 , wherein execution of the instructions further causes the computer system to:
automatically completing a portion of the computer programming source code based on the prediction of the programming element.

16. The computer system of claim 12 , wherein execution of the instructions further causes the computer system to:
in response to the application of the aggregated subset of machine learning models, detect an error in the computer programming source code.

17. The computer system of claim 16 , wherein execution of the instructions further causes the computer system to:
recommend one or more possible corrections of the error contained in a portion of the computer programming source code.

18. A method for providing custom machine learning models to a client computer system, the method comprising:
accessing a plurality of machine learning models;
accessing client-specific data for a client computer system;
using the client-specific data for the client computer system to determine which subset of the plurality of machine learning models is applicable to the client computer system, wherein:
the subset comprises multiple machine learning models selected from the plurality of machine learning models, and
determining which machine learning models are included in t

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Keech Jonathan Daniel
• Shanmugam Kesavan
• Calvert Simon
• Wilson-Thomas Mark A
• Lim Vivian Julia

Classifications

• G06F8/33

  Intelligent editors · CPC title

• G06N20/20 Primary

  Ensemble learning · CPC title

• G06F8/436

  Semantic checking · CPC title