LatentGAN Autoencoder: Learning Disentangled Latent Distribution

TL;DR

The paper introduces LatentGAN Autoencoder, a framework that learns and controls the autoencoder's latent distribution by training a LatentGAN generator to approximate latent space and a discriminator to distinguish real versus generated latent samples, augmented with a mutual-information objective to promote disentanglement. The approach combines autoencoder reconstruction with adversarial learning in the latent space, enabling direct manipulation of latent factors through discrete and continuous codes. Empirically, LatentGAN achieves a competitive $2.38\%$ unsupervised MNIST classification error, and demonstrates disentangled latent control on MNIST, 3D Chair, and CelebA datasets, outperforming InfoGAN and AAE baselines in key settings. This suggests that focusing on latent-space modeling rather than image-space generation can yield robust, controllable generative representations with practical benefits for unsupervised learning and conditional generation.

Abstract

In autoencoder, the encoder generally approximates the latent distribution over the dataset, and the decoder generates samples using this learned latent distribution. There is very little control over the latent vector as using the random latent vector for generation will lead to trivial outputs. This work tries to address this issue by using the LatentGAN generator to directly learn to approximate the latent distribution of the autoencoder and show meaningful results on MNIST, 3D Chair, and CelebA datasets, an additional information-theoretic constrain is used which successfully learns to control autoencoder latent distribution. With this, our model also achieves an error rate of 2.38 on MNIST unsupervised image classification, which is better as compared to InfoGAN and AAE.

Figures (8)

• Figure 1: LatentGAN Autoencoder model
• Figure 2: Varying $Cat(K)$ changes the digit category.Every row coressponds to different $K$, and every column have same $K$ but different Gaussian noise $P$.
• Figure 3: Varying $U_{1}(-1.5,1.5)$ changes the thickness of the digit.Every row have different $K$ and Gaussian noise $P$,and every column have different $U_{1}$ but same Gaussian noise $P$ and $K$.
• Figure 4: Varying $U_{2}(-1.5,1.5)$ rotates the digit.Every row have different $K$ and Gaussian noise $P$,and every column have different $U_{2}$ but same Gaussian noise $P$ and $K$.
• Figure 5: 3D Chair Generated Samples