Graph Convolution Node
Convolution performs a data operation using input matrix and bias data to produce an output matrix.
Filters, also called Kernels, remove unwanted data. During the forward pass, each filter uses a convolution process across the filter input, computing the dot product between the entries of the vectors in the filter filter and the input, producing a 2-dimensional output of that filter. As a result, through backpropagation the network learns filters that activate when it detects some specific type of feature at some spatial position in the input.
Below is an example of a graph segment depicting a simple convolution node:
where:
An example of calculating one output cell is: