Image Classification on MNIST (test) (Calibration Metrics)
95.3AccuracyRLA
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RLAHessian approximation type=low-rank, Backbone=LeNet, Number of parameter samples=1002026.05 | 95.3 | 0.609 | 0.009 | 0.042 | 0.472 | |
| DIMSHessian approximation type=low-rank, Backbone=LeNet, Number of parameter samples=100, friction coefficient η0=0.12026.05 | 95 | 0.271 | 0.008 | 0.035 | 0.379 | |
| LINLAHessian approximation type=low-rank, Backbone=LeNet, Number of parameter samples=1002026.05 | 86 | 2.608 | 0.026 | 0.127 | 0.508 | |
| LAHessian approximation type=low-rank, Backbone=LeNet, Number of parameter samples=1002026.05 | 56.4 | 11.502 | 0.083 | 0.408 | 0.614 |