Conv1D.cpp File Reference

Implements FlexNN::Layers::Conv1D forward/backward/propagate. More...

#include "layers/Conv1D.hpp"
#include <cassert>
#include <cmath>
#include "activations/detail.hpp"

Include dependency graph for Conv1D.cpp:

Namespaces
namespace	FlexNN
	Namespace for the FlexNN neural network library.
namespace	FlexNN::Layers

Detailed Description

Implements FlexNN::Layers::Conv1D forward/backward/propagate.

Forward uses im2col + single GEMM (Zflat = W * Col) so Eigen can vectorize/thread the hot math; padding is zero-fill at im2col time so the inner loop has no branch. Propagate (col2im) and grad share the same helper. Results match the naive loops to ~1e-12. The logic matches LLD_FLEXNN §6.2 and is verified by tests/conv1d_test.cpp via brute force and finite differences.