Shuhei Watanabe

Make discrete parameters filterable

1

# Feature Request In the trial table, I would like discrete parameters to be filterable like categorical parameters.

Visualize the parameter distributions of the top X% trials with density estimators

1

# Feature Request *Please write your suggestion here.* The concept by itself is the following:  I would like to have density estimators of each parameter with a slider so...

# Feature Request The current version does not show anything when the mouse hovers over these icons and I know that the hints will be shown when the three-line icon...

Typo?

1

https://github.com/automl/HPOBench/blob/master/hpobench/benchmarks/ml/tabular_benchmark.py#L166 Should it be: ```python cost_key = f"{evaluation}_costs" ``` ?

Somehow let users catch errors in function evaluations

1

Basically, this part squeezes errors caused in function evaluations, so it is hard to debug. https://github.com/automl/neps/blob/master/src/metahyper/api.py#L403-L451

poetry add neural-pipeline-search yields the maximum recursion depth error

1

```shell $ poetry add neural-pipeline-search==0.8.3 ``` Outcome: ```shell maximum recursion depth exceeded while calling a Python object ``` I have no clue. But at least, `pip install` works.

Hyperband bracket generation is inconsistent with the original Hyperband paper

1

[The HB paper](https://arxiv.org/abs/1603.06560) says that each SH bracket samples $n = \lceil \frac{s_{\max} + 1}{s + 1} \eta^s \rceil$ configurations; however, [this line](https://github.com/automl/neps/blob/master/src/neps/optimizers/multi_fidelity/successive_halving.py#L207) samples $n = \lfloor\lfloor \frac{s_{\max} + 1}{s...

Hyperband bracket generation is inconsistent with the original Hyperband paper

1

[The DEHB paper](https://arxiv.org/abs/2105.09821) says that each SH bracket samples $n = \lceil \frac{s_{\max} + 1}{s + 1} \eta^s \rceil$ configurations; however, [this line](https://github.com/automl/DEHB/blob/master/dehb/optimizers/dehb.py#L135) samples $n = \lfloor\lfloor \frac{s_{\max} + 1}{s...

Ambiguity in the query

3

This is not exactly an issue but leads to misunderstandings, so I would like to mention it here. 1. Which will be taken into account, `nepochs` in the argument of...

Termination in colorectal_histology due to memory overflow

3

As mentioned in the title, `colorectal_histology` is terminated while both `cifar10` and `fashion-mnist` work. It seems only `colorectal_histology` requires 16+GB RAM in loading the surrogate benchmark. Environment: - Ubuntu 18.04...