Explaining Neural Models for Image Classification
MTQ315 Seminar Course Project
Please see presentation.pdf
and report.pdf
for an introduction and further details.
To setup the environment use:
conda env create -f env.yaml
The code / experiments / results are available in jupyter notebooks of the form {METHOD}_{DATASET}.ipynb, where METHOD is shap
, cf
, and cfproto
for SHAP explanations, simple counterfactuals and prototype-guided counterfactuals respectively and {DATASET} is mnist
and fmnist
for MNIST and Fashion MNIST respectively.