This repository aims to replicate the results from the ICML2020 paper "Robust Learning with the Hilbert-Schmidt Independence Criterion"
- Paper: http://proceedings.mlr.press/v119/greenfeld20a/greenfeld20a-supp.pdf
- Original Implementation: https://github.com/danielgreenfeld3/XIC
Training Params:
- Batch Size: 32
- Learning Rate: 1e-3
- L2 Regularization (Weight Decay): 1e-3
- Num Epochs (Best of): 10
Experiment Params:
- Loss Function: HSIC
- Num. Trials: 20
Training Params:
- Batch Size: 32
- Learning Rate: 1e-3
- L2 Regularization (Weight Decay): None
- Num Epochs (Best of): 7
