Package klyn.math.nn.layers
Contains 12 type(s).
Types
Classes
- AveragePooling2D - Mean over each spatial window, excluding padded positions from its divisor.
- Conv2D - Channels-last two-dimensional cross-correlation with trainable kernel and bias.
- Conv2DTranspose - Trainable transposed convolution for spatial upsampling, not a mathematical inverse.
- Dense - An affine layer: output = inputs @ weights + bias, in row-major layout.
- DenseGradients - Gradients of an affine layer with respect to its inputs, weights and bias.
- Dropout - Random element-wise dropout during training, identity during inference.
- Flatten - Flattens sample axes while preserving the leading batch axis.
- Input - Declares and validates sample dimensions, excluding the batch axis.
- Layer - Statically typed stage of a sequential batch computation.
- LayerGradients - Explicit backward result, with parameter derivatives in layer parameter order.
- LayerPass - Explicit per-batch forward context, released by the trainer after backward.
- MaxPooling2D - Maximum over each spatial window, keeping channels independent.