Update rule learning on Conway's Game of Life minimal L(1, 1) models
100Success RatePolyKAN
Evaluation results
| Method | Links | |
|---|---|---|
| PolyKANTraining Protocol=Full parameters, Trainable Parameters=34, Monotonicity=strictly → non, Differentiability=smooth2026.06 | 100 | |
| PolyKANTraining Protocol=Only activation functions modifiable, Trainable Parameters=29, Monotonicity=strictly → non, Differentiability=smooth2026.06 | 100 | |
| PReLUTraining Protocol=Only activation functions modifiable, Trainable Parameters=23, Monotonicity=strictly → varies, Differentiability=discontinuous2026.06 | 100 | |
| PReLUTraining Protocol=Full parameters, Trainable Parameters=28, Monotonicity=strictly → varies, Differentiability=discontinuous2026.06 | 97 | |
| SquareTrainable Parameters=25, Monotonicity=non, Differentiability=smooth2026.06 | 94 | |
| SiLUTrainable Parameters=25, Monotonicity=non, Differentiability=smooth2026.06 | 94 | |
| PolyKANTraining Protocol=Only weights modifiable, Trainable Parameters=25, Monotonicity=strictly → non, Differentiability=smooth2026.06 | 78 | |
| PReLUTraining Protocol=Only weights modifiable, Trainable Parameters=25, Monotonicity=strictly → varies, Differentiability=discontinuous2026.06 | 59 | |
| RootSquareTrainable Parameters=25, Monotonicity=non, Differentiability=discontinuous2026.06 | 50 | |
| LeakyReLUTrainable Parameters=25, Monotonicity=strictly, Differentiability=discontinuous2026.06 | 25 | |
| CELUTrainable Parameters=25, Monotonicity=strictly, Differentiability=continuous2026.06 | 6 | |
| SigmoidTrainable Parameters=25, Monotonicity=strictly, Differentiability=smooth2026.06 | 0 | |
| TanhTrainable Parameters=25, Monotonicity=strictly, Differentiability=smooth2026.06 | 0 | |
| ReLUTrainable Parameters=25, Monotonicity=weakly, Differentiability=discontinuous2026.06 | 0 |