Bidirectional Conditional Generative Adversarial Networks
This work addresses a specific challenge in generative modeling for researchers and practitioners, offering incremental improvements in disentanglement and encoding accuracy.
The paper tackles the problem of disentangling latent variables and conditional information in conditional Generative Adversarial Networks (cGANs) by proposing Bidirectional cGAN (BiCoGAN), which includes an encoder for inverse mappings and training techniques, resulting in more accurate encoding and effective sample generation.
Conditional Generative Adversarial Networks (cGANs) are generative models that can produce data samples ($x$) conditioned on both latent variables ($z$) and known auxiliary information ($c$). We propose the Bidirectional cGAN (BiCoGAN), which effectively disentangles $z$ and $c$ in the generation process and provides an encoder that learns inverse mappings from $x$ to both $z$ and $c$, trained jointly with the generator and the discriminator. We present crucial techniques for training BiCoGANs, which involve an extrinsic factor loss along with an associated dynamically-tuned importance weight. As compared to other encoder-based cGANs, BiCoGANs encode $c$ more accurately, and utilize $z$ and $c$ more effectively and in a more disentangled way to generate samples.