Discriminative template learning in group-convolutional networks for invariant speech representations