Quantization-Based Regularization for Autoencoders
This work addresses regularization challenges in autoencoders for unsupervised representation learning, but it is incremental as it combines existing methods like VQ-VAE and denoising regularization.
The paper tackles the problem of overfitting and posterior collapse in autoencoders by introducing a quantization-based regularizer in the bottleneck stage, resulting in improved latent representations for supervised learning and clustering tasks compared to other bottleneck structures.
Autoencoders and their variations provide unsupervised models for learning low-dimensional representations for downstream tasks. Without proper regularization, autoencoder models are susceptible to the overfitting problem and the so-called posterior collapse phenomenon. In this paper, we introduce a quantization-based regularizer in the bottleneck stage of autoencoder models to learn meaningful latent representations. We combine both perspectives of Vector Quantized-Variational AutoEncoders (VQ-VAE) and classical denoising regularization methods of neural networks. We interpret quantizers as regularizers that constrain latent representations while fostering a similarity-preserving mapping at the encoder. Before quantization, we impose noise on the latent codes and use a Bayesian estimator to optimize the quantizer-based representation. The introduced bottleneck Bayesian estimator outputs the posterior mean of the centroids to the decoder, and thus, is performing soft quantization of the noisy latent codes. We show that our proposed regularization method results in improved latent representations for both supervised learning and clustering downstream tasks when compared to autoencoders using other bottleneck structures.