Table 6 Accuracy comparison of MNIST classification.
From: Deep representation learning using layer-wise VICReg losses
Num. of sample | Linear (%) | Baseline (%) | Fine-tuned (%) |
|---|---|---|---|
500 | 73.14 | 77.78 | 85.13 |
1000 | 74.75 | 81.33 | 88.66 |
2500 | 77.52 | 88.52 | 91.69 |
5000 | 78.50 | 93.68 | 94.23 |
