CMA stop conditions #401
Open
CMA stop conditions #401
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
|
Thanks for opening your first pull request in this repository! Someone will review it when they have a chance. In the mean time, please be sure that you've handled the following things, to make the review process quicker and easier:
Thank you again for your contributions! 👍 |
zoq
reviewed
Jun 16, 2024
| @@ -274,6 +291,17 @@ typename MatType::elem_type CMAES<SelectionPolicyType, | |||
| return overallObjective; | |||
| } | |||
|
|
|||
| // Terminate if sigma/sigma0 is above 10^4 * max(D) | |||
There was a problem hiding this comment.
Suggested change
| // Terminate if sigma/sigma0 is above 10^4 * max(D) | |
| // Terminate if sigma/sigma0 is above 10^4 * max(D). |
To be consistent with the style.
I think perhaps some review would be fine, but I believe its okay. It should be ready to be merged |
|
Since this is so close, we should keep it open to get it merged. I marked it keep-open |
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Outdated suggestions cannot be applied.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.
Hi,
I am currently enhancing the implementation of the CMA-ES algorithm by introducing additional stopping conditions. These modifications are intended to improve numerical stability and are advocated in A Restart CMA Evolution Strategy With Increasing Population Size and The CMA Evolution Strategy: A Tutorial. The changes are also in line with implementations found in Python libraries maintained by the same author.
These are basically:
maxiter): Stop if the number of iterations reaches a predefined limit.ftarget): Stop if the best objective value is less than or equal to a target value.maxfevals): Stop if the number of function evaluations reaches a predefined limit.tolfun): Stop if the range of function values of the best solutions falls below a threshold over a certain number of iterations.tolflatfitness): Stop if the best objective value does not change significantly over a number of generations.tolconditioncov): Stop if the condition number of the covariance matrix exceeds a certain threshold.noeffectaxis): Stop if adding 0.1 times the standard deviation in the search space in a principal axis direction does not change the solution.noeffectcoord): Stop if adding 0.2 times the standard deviation in the search space along one coordinate does not change the solution.tolx): Stop if the change in the variables between generations falls below a threshold.tolfacupx): Stop if any scaled step sizes exceed the product of the initial step sizes and a specified tolerance factor.tolfunhist): Stop if the range of fitness values in the historical best fitness list remains below a specified threshold.tolupsigma): Stop if the ratio of the current step size to the maximum eigenvalue of the covariance matrix exceeds a specified multiple of the initial step size.tolstagnation): Stop if there is no significant improvement over a large number of generations.tolfunrel): Stop based on a relative condition involving the historical and current best function values.These conditions are the ones currently present in the Python repository I mentioned and align with the papers while adding a few other stopping criteria. These are what I plan to implement, but I'm open suggestions.
Additionally, I have made modifications in lines 94 and 112 to align with the initialization presented in The CMA Evolution Strategy: A Comparing Review (page 26, equations 35 and 37). These initialization parameters are also supported by other papers I have reviewed.