def aggregate
neigh_dims = [batch_size * support_sizes[hop],
num_samples[len(num_samples) - hop - 1],
dim_mult*dims[layer]]
- support_size: 表示每一次sample的每一个点的总结点数
- batch_size: the size of each batch
reshape the hidden embedding of hop+1
layer to:
the number of layer hop * the number of neighbours of each node * the length of feather