Exploiting Pre-trained Feature Networks for Generative Adversarial Networks in Audio-domain Loop Generation.