By Ishita Mediratta, Snehanshu Saha, Shubhad Mathur.
Official implementation of "LipARELU: A-ReLU Networks aided by Lipschitz Acceleration", accepted to IJCNN 2021 (Oral).
For each task, either classification or regression, go to its corresponding subfolder. To train a model with a given loss type, run the entire jupyter notebook.
Loss types supported are:
| Classification | Regression | |
|---|---|---|
| MAE | ✅ | ✅ |
| Binary/Cross-Entropy | ✅ | ⛔️ |
| Quadratic Loss | ⛔️ | ✅ |