At this reading group we discussed a proposed alternative to existing sampling methods based on a fully connected neural network. After discussing the paper we concluded that some explanation is lacking and more details of the implementation would improve the paper.

Abstract

Accurate noise modelling is important for training of deep learning reconstruction algorithms. While noise models are well known for traditional imaging techniques, the noise distribution of a novel sensor may be difficult to determine a priori. Therefore, we propose learning arbitrary noise distributions. To do so, this paper proposes a fully connected neural network model to map samples from a uniform distribution to samples of any explicitly known probability density function. During the training, the Jensen-Shannon divergence between the distribution of the model’s output and the target distribution is minimized. We experimentally demonstrate that our model converges towards the desired state. It provides an alternative to existing sampling methods such as inversion sampling, rejection sampling, Gaussian mixture models and Markov-Chain-Monte-Carlo. Our model has high sampling efficiency and is easily applied to any probability distribution, without the need of further analytical or numerical calculations.

Link to the paper.

One of the references was interesting, it proposes a version of MCMC that incorporates a generative adversarial network (GAN). It may be worth discussing at a following meeting.