Generative Adversarial Stacked Autoencoders
This work offers an incremental improvement in the training stability and image quality for researchers and practitioners working with Adversarial Autoencoders.
This paper addresses the training difficulties of Generative Adversarial Networks (GANs), such as vanishing gradients and mode collapse. It proposes a novel Generative Adversarial Stacked Convolutional Autoencoder (GASCA) model and a gradual greedy layer-wise learning algorithm, which results in images with significantly lower reconstruction error compared to vanilla joint training.
Generative Adversarial Networks (GANs) have become predominant in image generation tasks. Their success is attributed to the training regime which employs two models: a generator G and discriminator D that compete in a minimax zero sum game. Nonetheless, GANs are difficult to train due to their sensitivity to hyperparameter and parameter initialisation, which often leads to vanishing gradients, non-convergence, or mode collapse, where the generator is unable to create samples with different variations. In this work, we propose a novel Generative Adversarial Stacked Convolutional Autoencoder(GASCA) model and a generative adversarial gradual greedy layer-wise learning algorithm de-signed to train Adversarial Autoencoders in an efficient and incremental manner. Our training approach produces images with significantly lower reconstruction error than vanilla joint training.