|
FlexNN 1
Fully connected neural network built from scratch with flexible n-layer design and multiple activations.
|
Parameters for BatchNorm1D. More...
#include <LayerTypes.hpp>
Public Attributes | |
| int | numFeatures = 0 |
| float | epsilon = 1e-5f |
| float | momentum = 0.1f |
Parameters for BatchNorm1D.
Epsilon is fixed at 1e-5 in v0.1 and not serialized; momentum is also fixed at 0.1 for running stats update. Future versions may serialize epsilon via aux if needed.
| float FlexNN::Layers::BatchNormParams::epsilon = 1e-5f |
| float FlexNN::Layers::BatchNormParams::momentum = 0.1f |
| int FlexNN::Layers::BatchNormParams::numFeatures = 0 |