Skip to content

Add/expose random seed in LearningPipeline to get deterministic results #9

Closed

Description

@GalOshri

It is important to be able to set a random seed for ML experiments so the results are reproducible. Add random seed to LearningPipeline (or elsewhere) to ensure a full experiment is deterministic.

Activity

  1. markusweimer commented on May 4, 2018

    @markusweimer

    Which of those variants do you prefer?

      // Variant 1: Constructor
      var p = new LearningPipeline(seed:1234);
    
      // Variant 2: Property
      p.seed = 1234;
    
      // Variant 3: Parameter to .Train()
      var m = p.Train(seed:1234);
  2. added this to the 0518 milestone on May 9, 2018
  3. zeahmed commented on May 11, 2018

    @zeahmed
    Contributor

    I'll prefer

    // Variant 3: Parameter to .Train()
    var m = p.Train(seed:1234);
  4. TomFinley commented on May 11, 2018

    @TomFinley
    Contributor

    Seriously, enough is enough I think. Let's just expose IHostEnvironment. That's our mechanism to solve this and other problems, it's worked well for years doing exactly this among all its other functions.

  5. modified the milestones: 0518, 0618 on May 22, 2018
  6. shauheen commented on Jun 28, 2018

    @shauheen
    Contributor

    With a few PRs already merged as well as the proposed #371 , this issue can now be closed.

  7. ghost locked as resolved and limited conversation to collaborators on Mar 31, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions