Autoencoder における中間層出力の分布に注目した評価基準の検討

抄録

Autoencoders that acquire specific feature space models from unsupervised data have become an important technique for designing systems based on neural networks. In this paper, we investigate of the features of high reusable autoencoder. We confirmed this ability from two experiments, focusing on the role reusable auto encoders. As a result, we found that the role that the intermediate units of the most reusable autoenocder in our experiments plays is regarded as binary function.