COORDINATED VERSION CONTROL SYSTEM, METHOD, AND RECORDING MEDIUM FOR PARAMETER SENSITIVE APPLICATIONS

• US 20170344905A1
• Filed: 05/31/2016
• Published: 11/30/2017
• Est. Priority Date: 05/31/2016
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Hack, Michel H.T., Ren, Yufei, Wang, Yandong, Zhang, Li
• Granted Patent: US 10,643,147 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/1873 Versioning file systems, temporal file systems, e.g. file system supporting different historic versions of files
• CPC Class Codes: G06F 16/219 Managing data history or versioning querying versioned data G06F16/2474; querying temporal data G06F16/2477
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Version vector-based rules are used to facilitate asynchronous execution of machine learning algorithms. The method uses version vector based rule to generate aggregated parameters and determine when to return the parameters. The method also includes coordinating the versions of aggregated parameter sets among all the parameter servers. This allows to broadcast to enforce the version consistency; generate parameter sets in an on-demand manner to facilitate version control. Furthermore the method includes enhancing the version consistency at the learner'"'"'s side and resolving the inconsistent version when mismatching versions are detected.

Claims

1. A version enforcement system including a parameter server having a plurality of parameter sets and a plurality of learners communicating with the plurality of parameter sets, the system comprising:

a parameter collecting circuit configured to collect the plurality of parameter sets from the parameter server;

a version examining circuit configured to determine versions of each of the parameter sets;

a mismatch detecting circuit configured to detect if a mismatch exists between the versions of each of the parameter sets;

a determining circuit configured to determine a correct version for each of the parameter sets to generate if the mismatch detecting circuit detects the mismatch;

a requesting circuit configured to request that said each of the parameter sets generates the correct version;

a success checking circuit configured to verify that said each of the parameter sets generates the correct version in response to the request by the requesting circuit; and

a local parameter updating circuit configured to automatically update local parameters for computation by a learner with the correct version of the aggregated parameter sets.

2. The system of claim 1, further comprising a retrying circuit configured to cause the parameter sets to retry to generate the correct version for a predetermined number of attempts or for a predetermined amount of time.

3. The system of claim 2, further comprising a time-out determining circuit configured to determine that the correct version has not been generated within the predetermined number of attempts or the predetermined amount of time, and

wherein the time-out determining circuit causes the parameter collecting circuit to re-collect the plurality of parameter sets from the parameter server such that the determining circuit can determine a new correct version.

4. The system of claim 1, wherein, if the mismatch detecting circuit detects that there is not a mismatch between versions of each of the parameter sets, then the local parameter updating circuit automatically updates local parameters and causes computation by a learner for the matching versions of each of the parameter sets.

5. The system of claim 1, wherein the determining circuit uses a Quorum approach to determine which version to follow as the correct version.

6. The system of claim 1, wherein the correct version includes a most recent version of a computation from each of the plurality of learners.

7. The system of claim 1, wherein the correct version includes a version currently existing in at least one of the parameter sets.

8. The system of claim 1, wherein each of the parameter sets saves previous versions of the parameter sets.

9. The system of claim 1, wherein if the parameter set of the parameter sets receives a request from the requesting circuit to generate an older version of the parameter set, the parameter set uses a cached version to generate a particular version to match the version of the correct version.

18. The method of claim 1, wherein if the parameter set of the parameter sets receives a request from the requesting to generate an older version of the parameter set, the parameter set uses a cached version to generate a particular version to match the version of the correct version.

10. A version enforcement method including a parameter server having a plurality of parameter sets and a plurality of learners communicating with the plurality of parameter sets, the method comprising:

collecting the plurality of parameter sets from the parameter server;

determining versions of each of the parameter sets;

detecting if a mismatch exists between the versions of each of the parameter sets;

determining a correct version for each of the parameter sets to generate if the detecting detects the mismatch;

requesting that said each of the parameter sets generate the correct version;

checking that said each of the parameter sets generate the correct version in response to the request by the requesting; and

automatically updating local parameters for computation by a learner with the correct version of the aggregated parameter sets.

11. The method of claim 10, further comprising causing the parameter sets to retry to generate the correct version for a predetermined number of attempts or for a predetermined amount of time.

12. The method of claim 11, further comprising determining a time-out by determining that the correct version has not been generated within the predetermined number of attempts or the predetermined amount of time, and

wherein determining the time-out causes the collecting to re-collect the plurality of parameter sets from the parameter server such that the determining can determine a new correct version.

13. The method of claim 10, wherein, if the mismatch detecting detects that there is not a mismatch between versions of each of the parameter sets, then the local parameter updating automatically updates local parameters and causes computation by a learner for the matching versions of each of the parameter sets.

14. The method of claim 10, wherein the determining uses a Quorum approach to determine which version to follow as the correct version.

15. The method of claim 10, wherein the correct version includes a most recent version of a computation from each of the plurality of learners.

16. The method of claim 10, wherein the correct version includes a version currently existing in at least one of the parameter sets.

17. The method of claim 10, wherein each of the parameter sets saves previous versions of the parameter sets.

19. An asynchronous execution facilitation system including a parameter server having a plurality of parameter sets and a plurality of learners communicating with the plurality of parameter sets, comprising:

a server checking circuit configured to check each pull request to the parameter server for an immediate version vector and a local epoch of the learner; and

an aggregation returning circuit configured to compare the immediate version vector pulled by the parameter server with a comparison condition such that if the comparison condition is satisfied, the aggregation returning circuit causes the parameter server to return a new aggregated parameter to the learner, the comparison condition being a comparison of the immediate version vector to a current version of the aggregated parameter set of the parameter server.

20. A non-transitory computer-readable recording medium recording for version enforcement performing the method of claim 19.