Add x log scale for slice plot

[keisuke-umezawa]

Contributor License Agreement

This repository (optuna-dashboard) and Goptuna share common code.
This pull request may therefore be ported to Goptuna.
Make sure that you understand the consequences concerning licenses and check the box below if you accept the term before creating this pull request.

• I agree this patch may be ported to Goptuna by other Goptuna contributors.

Reference Issues/PRs

To fix the #64

What does this implement/fix? Explain your changes.

Add switch for the log scale of x axis in addition to y axis.

[keisuke-umezawa]

If you have a better name than Log x scale, please tell me that.

[c-bata]

Thank you for your pull request.
Let me assign @chenghuzi as a reviewer.

[chenghuzi]

LGTM! Just one thing: sometimes I need to click the switch twice to turn the log scale off or on. I noticed that we have the same issue in the log scale switch in the HistoryGraph. So probably we should remove the line e.preventDefault() in handleChange functions? It seems in the official doc this is also not necessary.

[c-bata]

I think it's better to automatically determine x axis from Optuna's distribution instead of the toggle button. What do you think?

[chenghuzi]

I think it's better to automatically determine x axis from Optuna's distribution instead of the toggle button. What do you think?

Agreed. Could we just add an association between x axis scale and the type of distribution so that the graph can switch scale automatically?

[c-bata]

Could we just add an association between x axis scale and the type of distribution so that the graph can switch scale automatically?

Yes. We can associate x axis scale with the distribution from intersection_search_space or union_search_space in the StudyDetailResponse so that the toggle button is no longer needed.

optuna-dashboard/optuna_dashboard/static/apiClient.ts

Lines 45 to 46 in 9053ce9

	intersection_search_space: SearchSpace[]
	union_search_space: SearchSpace[]

[c-bata]

@keisuke-umezawa Can you work on it?

[keisuke-umezawa]

I will!

[jeremycochoy]

Would this works well if the search space is dynamic and the variable was sampled both with log and uniform distribution?

[c-bata]

Would this works well if the search space is dynamic and the variable was sampled both with log and uniform distribution?

Good point 👍 In such a case, I think it's enough just showing in linear scale.

[keisuke-umezawa]

@chenghuzi I completed implementation. Could you review it?

[chenghuzi]

Overall LGTM! Just one thing, probably we can change the label of the log scale switch to specific Y axis.

[chenghuzi]

Overall LGTM! Thanks for your PR @keisuke-umezawa! I'll merge and close this.