|
FlexNN 1
Fully connected neural network built from scratch with flexible n-layer design and multiple activations.
|
Sigmoid activation — 1/(1+exp(-Z)) with clamping. More...
#include <algorithm>#include <Eigen/Dense>#include <cmath>

Go to the source code of this file.
Namespaces | |
| namespace | FlexNN |
| Namespace for the FlexNN neural network library. | |
| namespace | FlexNN::Activations |
| namespace | FlexNN::Activations::detail |
Functions | |
| Eigen::MatrixXd | FlexNN::Activations::detail::sigmoid_forward (const Eigen::MatrixXd &Z) |
| Sigmoid forward: A = 1/(1+exp(-clamp(Z))) elementwise. | |
| Eigen::MatrixXd | FlexNN::Activations::detail::sigmoid_backward (const Eigen::MatrixXd &dA, const Eigen::MatrixXd &, const Eigen::MatrixXd &A) noexcept |
| Sigmoid backward: dZ = dA * A * (1 - A). | |
Sigmoid activation — 1/(1+exp(-Z)) with clamping.
Clamping Z to [-15,15] avoids exp overflow and keeps host double stable while the runtime may use a LUT. Matches LLD_FLEXNN §6.5.