Mainak Jas

not fully clear in my head what additional info needs to be added to RejectLog but I'm open to it if it's backwards compatible.

@adam2392 can you propose an API for this? It would be great to have `autoreject` work out of the box on raw data for TFR analysis etc but I don't...

The data interpolation won't be good if you have too few channels. It's a good question as to how many channels are necessary for the interpolation to work. But 3...

@TomDonoghue the notion of `fit` and `transform` is for such use cases. If you do: ```py ar.fit(epochs) ``` and then later ```py ar.transform(epochs_filt1) ar.transform(epochs_filt2) # etc ``` It should reuse...

I see what you mean and your code seems to be just fine for the use case you mention. Except that it uses private API. I suppose then you do...

Can you try `method='bayesian_optimization'` instead of `method='random_search'` to see if that works?

Nice!

@dengemann "bayesian optimization" works better than "random search": that is for sure: and it is the default too as far as I remember. We had "random search" in one of...

hey @adam2392 ! yes this is the right place to ask the questions What will you use these data for? I am assuming for some time-frequency analysis? In that case...

also reminds me of this issue: https://github.com/autoreject/autoreject/issues/127