Style Memory: Making a Classifier Network Generative
This work addresses the limitation of losing stylistic details in classification for AI systems, though it appears incremental as it builds on existing autoencoder and classifier architectures.
The paper tackles the problem of classifier networks discarding stylistic information by introducing a 'style memory' network that performs both classification and reconstruction, achieving good reconstructions when classification is correct.
Deep networks have shown great performance in classification tasks. However, the parameters learned by the classifier networks usually discard stylistic information of the input, in favour of information strictly relevant to classification. We introduce a network that has the capacity to do both classification and reconstruction by adding a "style memory" to the output layer of the network. We also show how to train such a neural network as a deep multi-layer autoencoder, jointly minimizing both classification and reconstruction losses. The generative capacity of our network demonstrates that the combination of style-memory neurons with the classifier neurons yield good reconstructions of the inputs when the classification is correct. We further investigate the nature of the style memory, and how it relates to composing digits and letters. Finally, we propose that this architecture enables the bidirectional flow of information used in predictive coding, and that such bidirectional networks can help mitigate against being fooled by ambiguous or adversarial input.