Unsupervised Training of Convex Regularizers using Maximum Likelihood Estimation

arXiv:2404.05445v1 Announce Type: cross
Abstract: Unsupervised learning is a training approach in the situation where ground truth data is unavailable, such as inverse imaging problems. We present an unsupervised Bayesian training approach to learning convex neural network regularizers using a fixed noisy dataset, based on a dual Markov chain estimation method. Compared to classical supervised adversarial regularization methods, where there is access to both clean images as well as unlimited to noisy copies, we demonstrate close performance on natural image …

abstract arxiv bayesian cs.lg data dataset imaging likelihood markov maximum likelihood estimation network neural network stat.co stat.me training truth type unsupervised unsupervised learning