Predicting Transformer Generalization on MNIST-Transformers No threshold
0.911Kendall's TauTransformer-NFN Quasi
Evaluation Results
| Method | Links | |
|---|---|---|
| Transformer-NFN QuasiExtra Parameters (%)=4.54% params ++, Number of runs=52026.04 | 0.911 | |
| Transformer-NFN largeExtra Parameters (%)=57.66% params ++, Number of runs=52026.04 | 0.907 | |
| Transformer-NFNNumber of runs=52026.04 | 0.905 | |
| STATNNNumber of runs=52026.04 | 0.881 | |
| MLPNumber of runs=52026.04 | 0.866 | |
| XGBoostNumber of runs=52026.04 | 0.86 | |
| LightGBMNumber of runs=52026.04 | 0.858 | |
| Random ForestNumber of runs=52026.04 | 0.772 |