- To implement Representation Learning algorithms (Contrastive and self-supervised learning algorithms)
- SimCLR (ICML 2020)
- SupCon (NIPS 2020)
- SimSiam (CVPR 2021)
- Bootstrap your own Latent (BYOL) (NIPS 2020)
- Barlow Twins (ICML 2021)
- TripletMarginLoss (CVPR 2015)
| Algorithm | CIFAR10 (R50) | CIFAR100 (R50) | CIFAR10 (R18) | CIFAR100 (R18) |
|---|---|---|---|---|
| SimCLR | 87.5 | 57.7 | 85.9 | 55.0 |
| SupCon | 94.0 | 74.7 | 93.5 | 70.4 |
| Triplet | 83.4 | 76.3 | 86.0 | 64.5 |
| Barlow Twins | 81.2 | 47.7 | 80.3 | 45.8 |
| BYOL | 83.0 | 47.0 | 84.8 | 54.8 |
| SimSiam | 76.5 | 34.5 | 88.6 | 62.3 |
| DARe (Ours) | 89.4 | 62.3 | 87.3 | 61.6 |
- Distribution Alignment Regularizer (DARe) contrastive learning
- Mansi Tomer (SR: 24014)
- Nirbhay Sharma (SR: 24806)