[dattri.benchmark] Add a new benchmark model MLP

TRAIS-Lab / dattri

`dattri` is a PyTorch library for developing, benchmarking, and deploying efficient data attribution algorithms.

https://trais-lab.github.io/dattri/

24 stars 8 forks source link

[dattri.benchmark] Add a new benchmark model MLP #70

Closed TheaperDeng closed 3 months ago

TheaperDeng commented 3 months ago

Description

1. Motivation and Context

Add a new benchmark model MLP, this model is currently working with MNIST 10 dataset.

2. Summary of the change

Same as MNIST + LR setting.

3. What tests have been added/updated for the change?

[x] Unit test: Typically, this should be included if you implemented a new function/fixed a bug.

TheaperDeng commented 3 months ago

TODO: check the accuracy Update 5/14: Checked.

IF mislabeled detection passed (find more than 86% mislabeled data after swap 20% of the dataset):

Checked Data Sample      Found flipped Sample
--------------------------------------------------
0                        0
100                      61
200                      86
300                      92
400                      96
500                      98
600                      99
700                      99
800                      99
900                      99

TheaperDeng commented 3 months ago

I will merge this PR first to keep rolling since it's only a follow-up work. I will fix the comments later.