Two-Dimensional Convolution Forward Layer
- Four-dimensional tensorX∈Rwith input datan1xn2xn3xn4
- Four-dimensional tensorK∈Rwith kernel parameters/weights of kernels (convolutions)nKernelsxm2xm3xm4
- One-dimensional tensorB∈Rwith the bias of each kernel.nKernels
- pis the respective padding.i
- nGroupsis defined as follows: let's assume thatn2is the group dimension. The input tensor is split along this dimension intonGroupsgroups, the tensors of values and weights are split intonGroupsgroups along thenKernelsdimension.nKernelsandn2must be multiples ofnGroups. Each group of values is computed using the respective group in tensors of input data, weights, and biases.
- s3ands4are strides