Skip to main content

conv1d_raw

Function conv1d_raw 

Source
pub fn conv1d_raw(
    signal: &[f64],
    filters: &[f64],
    bias: &[f64],
    out: &mut [f64],
    signal_len: usize,
    out_channels: usize,
    kernel_size: usize,
)
Expand description

1D convolution with stride=1, no padding (“valid” mode).

signal: input signal of length signal_len. filters: [out_channels, kernel_size] row-major. bias: per-channel bias of length out_channels. out: output buffer [out_channels, out_len], out_len = signal_len - kernel_size + 1. Uses Kahan summation for deterministic dot products.