Coresets for unsupervised active learning.

[Dref360]

Is your feature request related to a problem? Please describe.
Cold-start is a real problem in active learning, we need to label randomly a few dozens labels before starting using active learning.

Describe the solution you'd like
A solution similar to Baal AbtractHeuristic.

coreset_algo = MyCoresetAlgorithm()
ranks = coreset_algo(my_dataset)
# or 
ranks = coreset_algo(predictions)

Describe alternatives you've considered
None

[Dref360]

Reference: Active Learning for Convolutional Neural Networks: A Core-Set Approach, https://arxiv.org/abs/1708.00489

https://arxiv.org/pdf/1910.08707.pdf

[GeorgePearse]

Did this get anywhere? Probably the PR I'd be most keen to try to implement.

[Dref360]

No we didn't worked on this.

Yes that would be amazing if you could get the ball rolling on this. We will be there to help that's for sure!

[GeorgePearse]

The apricot package actually looks interesting for solving the cold start problem. https://apricot-select.readthedocs.io/en/latest/ (doesn't require labels).

Don't know if it'd make sense to test it and then add documentation on their combined use.

[nitish1295]

You might find this useful https://github.com/decile-team/cords