Add EPIG

[Dref360]

This could unlock some new opportunities for us, particularly for Transformers.

Github
Paper

[reeshipaul]

I would like to work on this implementation

[Dref360]

Awesome!
Thank you very much :)

We would be happy to help you as well if needed

[Dref360]

I'll read the full paper over the weekend. I would know more about how we could work on this.

[Dref360]

After reading the paper and looking at the codebase, I think the main issue is that we can't easily supply all targets to the heuristic.

We could directly provide the active dataset to the heuristic and then iterate over it when we call EPIG(active_set=active_set).get_scores(predictions).

What do you think? I don't think this is the perfect solution so I can be easily persuaded to go another route.

[reeshipaul]

I have read the paper, will read it in depth once more till the weekend. Can we schedule a meet over the weekend to discuss about the implementation?

[Dref360]

Sounds good!

My calendar is available on Calendly: https://calendly.com/baal-org/30min?month=2023-05

If these don't work for you let me know and we will figure something out :) We can chat on Slack as well.

[reeshipaul]

@Dref360 can you please send out an invite for slack to reeshipaul35@gmail.com

[Dref360]

Fixed in #270 thank you @reeshipaul for the PR!