Semi-Supervised Adversarial Variational Autoencoder

Abstract

We present a method to improve the reconstruction and generation performance of a variational autoencoder (VAE) by injecting an adversarial learning. Instead of comparing the reconstructed with the original data to calculate the reconstruction loss, we use a consistency principle for deep features. The main contributions are threefold. Firstly, our approach perfectly combines the two models, i.e., GAN and VAE, and thus improves the generation and reconstruction performance of the VAE. Secondly, the VAE training is done in two steps, which allows to dissociate the constraints used for the construction of the latent space on the one hand, and those used for the training of the decoder. By using this two-step learning process, our method can be more widely used in applications other than image processing. While training the encoder, the label information is integrated to better structure the latent space in a supervised way. The third contribution is to use the trained encoder for the consistency principle for deep features extracted from the hidden layers. We present experimental results to show that our method gives better performance than the original VAE. The results demonstrate that the adversarial constraints allow the decoder to generate images that are more authentic and realistic than the conventional VAE.

Deep Residual Learning-Based Convolutional Variational Autoencoder For Driver Fatigue Classification

“…As mentioned in section 2, utilizing CVAE as a solitary classification method is fruitless because of the autoencoder is only tasked with regenerating input [18]. In order to implement discriminative modeling with the help of a deep generative model CVAE, the authors of this article have used a semi-supervised learning approach based on the study done in the article [19].…”

“…authors have investigated the effects of α and β on the testing performance of the network. According to[19], the best result has been obtained by the combination of values of α = 1 and β = 0.1. Authors have tested this combination against the identical combination (i.e.…”

The Utility of Deep Learning in Breast Ultrasonic Imaging: A Review

“…VAE differs from AE in that it provides a statistical manner to describe the samples of the dataset in latent space. Therefore, in VAE, the encoder outputs the probability distribution to the bottleneck layer instead of a single output value [57].…”

ECM-LSE: Prediction of Extracellular Matrix Proteins Using Deep Latent Space Encoding of k-Spaced Amino Acid Pairs

“…In the recent past, inspired by the success of deep long short-term memory (LSTM) models, some approaches similar to word2vec ( Mikolov et al, 2013 ) have been proposed to successfully learn latent space encoding directly from variable length sequences ( Ding et al, 2019 ). The direct sequence to latent space encoding method produces good generalization models ( Zemouri, 2020 ); however, they usually rely on the availability of a large training dataset. Furthermore, the direct extraction of latent space features from a limited number of sequences such as, bioluminescence ( Zhang et al, 2021 ), antioxidant ( Olsen et al, 2020 ), ECM ( Kabir et al, 2018 ), antifreeze proteins (AFPs) ( Kandaswamy et al, 2011 ), or other classes of proteins is a challenging problem.…”