SHOT-VAE: Semi-supervised Deep Generative Models With Label-aware ELBO Approximations
This work improves the accuracy of semi-supervised variational autoencoders by addressing specific limitations in their objective function, which is an incremental improvement for researchers and practitioners using VAEs.
This paper addresses the issue in semi-supervised VAEs where high ELBO values do not always correlate with accurate inference. The authors propose SHOT-VAE, which introduces a smooth-ELBO that integrates label predictive loss and an optimal interpolation approximation to overcome an ELBO value bottleneck. SHOT-VAE achieves a 25.30% error rate on CIFAR-100 with 10k labels and a 6.11% error rate on CIFAR-10 with 4k labels.
Semi-supervised variational autoencoders (VAEs) have obtained strong results, but have also encountered the challenge that good ELBO values do not always imply accurate inference results. In this paper, we investigate and propose two causes of this problem: (1) The ELBO objective cannot utilize the label information directly. (2) A bottleneck value exists and continuing to optimize ELBO after this value will not improve inference accuracy. On the basis of the experiment results, we propose SHOT-VAE to address these problems without introducing additional prior knowledge. The SHOT-VAE offers two contributions: (1) A new ELBO approximation named smooth-ELBO that integrates the label predictive loss into ELBO. (2) An approximation based on optimal interpolation that breaks the ELBO value bottleneck by reducing the margin between ELBO and the data likelihood. The SHOT-VAE achieves good performance with a 25.30% error rate on CIFAR-100 with 10k labels and reduces the error rate to 6.11% on CIFAR-10 with 4k labels.