Anes Benmerzoug

In order to prepare for a newer interface for game theoretic data valuation methods i.e. #467, we should deprecate the [`compute_shapley_values`](https://github.com/aai-institute/pyDVL/blob/96326ae75600391e927b01bf2e31a93f12915159/src/pydvl/value/shapley/common.py) and [`compute_least_core_values`](https://github.com/aai-institute/pyDVL/blob/96326ae75600391e927b01bf2e31a93f12915159/src/pydvl/value/least_core/__init__.py#L42) functions.

Implement Variance Reduced Data Shapley (VRDS)

2

Introduced in Wu, M., Jia, R., Huang, W., & Chang, X. (2022). [Robust Data Valuation via Variance Reduced Data Shapley](https://arxiv.org/abs/2210.16835). arXiv preprint arXiv:2210.16835. The idea is to use stratified sampling...

Move tolerate fixture to separate repository

2

With the changes in #529, we no longer use the tolerate fixture and we should therefore remove it from the repository. However I think it is useful and there may...

Issue with SemiValue batching and parallelization

5

While working on PR #341, I realized that there is a bug in the batching feature of semivalues when using `n_jobs` > 1. The results are almost the same but...

In order to show that data valuation is not limited to scikit-learn models, we should create a notebook demonstrating the use of a PyTorch with one of the data valuation...

Improve `Dataset` class

4

The current `Dataset` implementation is very limited and does not work nicely with PyTorch, for example. We should make it closer to the implementation and have a better separation between...

As a follow-up to #189, We should create a notebook showing how to use caching and when it makes sense to use it and when it doesn't (e.g. Permutation vs...