Image Classification on CIFAR-10 (val) (Accuracy and Training Success Rate)
81.96Average Validation AccuracyHyperbolic
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HyperbolicMirror potential=Hyperbolic, Backbone=VGG16, Batch Normalization=None, Weight Decay=None, Momentum=None2026.05 | 81.96 | — | — | |
| GDMirror potential=GD, Backbone=VGG16, Batch Normalization=None, Weight Decay=None, Momentum=None2026.05 | 81.26 | — | — | |
| p = 10Mirror potential=p = 10, Backbone=VGG16, Batch Normalization=None, Weight Decay=None, Momentum=None2026.05 | 81.26 | — | — | |
| FractalNetMean training time per epoch=~5 minutes, Mean GPU memory consumption=4–5 GB2025.11 | 60 | 80.18 | 97 |