It Takes (Only) Two: Adversarial Generator-Encoder Networks
This work addresses the challenge of enhancing unsupervised learning for generative models, offering a simpler and more efficient approach for researchers and practitioners in machine learning, though it is incremental relative to existing hybrid methods.
The paper tackles the problem of improving sample quality in autoencoder-based generative models by introducing an adversarial game directly between the encoder and generator, without external mappings. The result is a model that achieves comparable quality in samples and reconstructions to more complex architectures, as demonstrated in experiments.
We present a new autoencoder-type architecture that is trainable in an unsupervised mode, sustains both generation and inference, and has the quality of conditional and unconditional samples boosted by adversarial learning. Unlike previous hybrids of autoencoders and adversarial networks, the adversarial game in our approach is set up directly between the encoder and the generator, and no external mappings are trained in the process of learning. The game objective compares the divergences of each of the real and the generated data distributions with the prior distribution in the latent space. We show that direct generator-vs-encoder game leads to a tight coupling of the two components, resulting in samples and reconstructions of a comparable quality to some recently-proposed more complex architectures.