2020/06/18 by J. Deasy, Deasy, Jacob, Nikola Simidjievski +3 · 4 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · #Anomaly Detection Techniques and Applications #Cell Image Analysis Techniques #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG) #Machine Learning (stat.ML)
paper · pdf · doi:10.48550/arxiv.2006.10599
openalex publication_date 2020/06/18 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
We examine the problem of controlling divergences for latent space\nregularisation in variational autoencoders. Specifically, when aiming to\nreconstruct example x\∈\ℝm via latent space z\∈\ℝn\n(n\≤ m), while balancing this against the need for generalisable latent\nrepresentations. We present a regularisation mechanism based on the\nskew-geometric Jensen-Shannon divergence\n\( textrmJS^ textrmG\α\). We find a variation in\n textrmJS^ textrmG\α, motivated by limiting cases, which leads\nto an intuitive interpolation between forward and reverse KL in the space of\nboth distributions and divergences. We motivate its potential benefits for VAEs\nthrough low-dimensional examples, before presenting quantitative and\nqualitative results. Our experiments demonstrate that skewing our variant of\n textrmJS^ textrmG\α, in the context of\n textrmJS^ textrmG\α-VAEs, leads to better reconstruction and\ngeneration when compared to several baseline VAEs. Our approach is entirely\nunsupervised and utilises only one hyperparameter which can be easily\ninterpreted in latent space.\n