Project DetailsGenerative models have been widely studied in recent years using large and costly DNN-based models. Yet, results still have much room for improvement in terms of both accuracy and runtime. In our work, we aim to tackle the Generative Modeling problem using a different, computationally lighter approach, based on an iterative fitting process between marginals of source and target distributions. Intuitively, one can think of this process as an analogue to the process of Tomography where each direction of observation adds information of the objects density. In this report, we formulate the underlying theory, demonstrate the algorithms performance, and analyze its abilities and weaknesses. Additionally, we extend our architecture to fit high-dimensional distributions (MNIST, CIFAR-10). Our results are not competitive with those of existing state-of-the-art methods, yet we show some advantages over traditional methods and believe that our approach has its potential for future work.
