catcma (GECCO2024) #178
catcma (GECCO2024) #178
Conversation
| _SIGMA_MAX = 1e32 | ||
|
|
||
|
|
||
| class CatCMA: |
There was a problem hiding this comment.
Can you add the link for the paper?
e.g. https://github.com/CyberAgentAILab/cmaes/blob/main/cmaes/_cmawm.py#L98
There was a problem hiding this comment.
I added the link for the paper (arXiv).
cmaes/_catcma.py
Outdated
| min_eigenvalue: | ||
| Lower bound of eigenvalue of multivariate Gaussian distribution (optional). | ||
|
|
||
| delta_init: |
There was a problem hiding this comment.
I'm not sure whether there is a situation such that we want to tune this.
There was a problem hiding this comment.
The variables delta_init and Delta_max have been removed from the __init__ method.
cmaes/_catcma.py
Outdated
| delta_init: | ||
| Initial trust region radius in ASNG (optional). | ||
|
|
||
| Delta_max: |
There was a problem hiding this comment.
The variables delta_init and Delta_max have been removed from the __init__ method (as mentioned above).
| seed: Optional[int] = None, | ||
| population_size: Optional[int] = None, | ||
| cov: Optional[np.ndarray] = None, | ||
| cat_param: Optional[np.ndarray] = None, |
There was a problem hiding this comment.
Please add the validation.
There was a problem hiding this comment.
I added the validation for self._n_ca and cat_param.
|
Thanks! LGTM. |
I have implemented CatCMA [Hamano+, GECCO2024].