Generative adversarial networks (GANs) learn the distribution of observed samples through a zero-sum game between two machine players, generator and discriminator. While GANs achieve great success in learning complex image, sound, text data, they perform suboptimally multimodal distribution-learning benchmarks such as Gaussian mixture models (GMMs). In this paper, we propose Adversarial Trainin...

Abstract This paper studies how to learn parameters in diagonal Gaussian mixture models. The problem can be formulated as computing incomplete symmetric tensor decompositions. We use generating polynomials compute decompositions and approximations. Then the approximation method is used also do stability analysis. When first third order moments are sufficiently accurate, we show that obtained fo...