2021/11/24 by Ankita Shukla, Shukla, Ankita, Rushil Anirudh +13
Engineering · Computer Science · #Particle Dynamics in Fluid Flows #Generative Adversarial Networks and Image Synthesis #Anomaly Detection Techniques and Applications
paper · pdf · doi:10.48550/arxiv.2111.12798
In this paper, we develop a Wasserstein autoencoder (WAE) with a\nhyperspherical prior for multimodal data in the application of inertial\nconfinement fusion. Unlike a typical hyperspherical generative model that\nrequires computationally inefficient sampling from distributions like the von\nMis Fisher, we sample from a normal distribution followed by a projection layer\nbefore the generator. Finally, to determine the validity of the generated\nsamples, we exploit a known relationship between the modalities in the dataset\nas a scientific constraint, and study different properties of the proposed\nmodel.\n