Reset random state on seed change #2130
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Jan Vesely <[email protected]>
… of RandomState Signed-off-by: Jan Vesely <[email protected]>
…function instead of RandomState Results change since the 'seed' parameter is converted to list before use. Signed-off-by: Jan Vesely <[email protected]>
Drop custom in-flight function check. Remove any cost calculation and use generated random value as objective metric. Signed-off-by: Jan Vesely <[email protected]>
…ialize parameters.random_state Fix function copy workaround to reset the seed parameter, the state will be adjusted by the seed setter function. Use DEFAULT_SEED to avoid advancing global seed. Signed-off-by: Jan Vesely <[email protected]>
|
This PR causes the following changes to the html docs (ubuntu-latest-3.7-x64): See CI logs for the full diff. |
|
This pull request fixes 1 alert when merging b0898ce into aea0621 - view on LGTM.com fixed alerts:
|
Instead of resetting the PRNG state once per context, the new approach compares the originally used seed with the new value (either new setting or modulation). This allows for better match in multi-trial executions with OCM: Simulations will only reset PRNG if the modulated seed is different from the one in the current execution context. The result of evaluation will thus match the result of trial even if OCM selects the same seed as originally used by the execution. The PRNG state is also reset on direct seed modification to preserve the old behaviour in situations like; Func1.seed = n Func2.seed = n # Both Func1 and Func2 now follow the same random sequence irrespective # of their original seed Signed-off-by: Jan Vesely <[email protected]>
…ials Construct the expected values in the test instead of using constants to verify the semantics of reusing random seeds. Signed-off-by: Jan Vesely <[email protected]>
|
This PR causes the following changes to the html docs (ubuntu-latest-3.7-x64): See CI logs for the full diff. |
|
This pull request fixes 1 alert when merging f32638b into aea0621 - view on LGTM.com fixed alerts:
|
kmantel
approved these changes
Oct 1, 2021
|
This PR causes the following changes to the html docs (ubuntu-latest-3.7-x64): See CI logs for the full diff. |
|
This pull request fixes 1 alert when merging 429ffea into 2c3a429 - view on LGTM.com fixed alerts:
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The original semantics was to reset random state once per context with the most up-to-date value of the seed.
The new approach is to reset random state every time the seed value changes, whether through direct assignment or modulation.
The practical change in semantics is subtle and helps match results between virtual executions and the simulated composition (see the comments in 'test_modulation_of_random_state' for details).