Update README.md
Browse files
README.md
CHANGED
|
@@ -13,6 +13,7 @@ MobiusNet introduces a fundamentally different approach to neural network design
|
|
| 13 |
- **Thirds Mask**: Cantor-inspired fractal channel suppression for regularization
|
| 14 |
- **Continuous Topology**: Layers sample a continuous manifold via the `t` parameter, not discrete units
|
| 15 |
- **Twist Rotations**: Smooth rotation through representation space across network depth
|
|
|
|
| 16 |
|
| 17 |
## Performance
|
| 18 |
|
|
|
|
| 13 |
- **Thirds Mask**: Cantor-inspired fractal channel suppression for regularization
|
| 14 |
- **Continuous Topology**: Layers sample a continuous manifold via the `t` parameter, not discrete units
|
| 15 |
- **Twist Rotations**: Smooth rotation through representation space across network depth
|
| 16 |
+
- **Integrator**: The integrator uses GELU in experimentation to enable additional GELU-based nonlinearity.
|
| 17 |
|
| 18 |
## Performance
|
| 19 |
|