Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer