Learning MMSE Optimal Thresholds for FISTA

Fast iterative shrinkage/thresholding algorithm (FISTA) is one of the most commonly used methods for solving linear inverse problems. In this work, we present a scheme that enables learning of optimal thresholding functions for FISTA from a set of training data. In particular, by relating iterations of FISTA to a deep neural network (DNN), we use the error backpropagation algorithm to find thresholding functions that minimize mean squared error (MSE) of the reconstruction for a given statistical distribution of data. Accordingly, the scheme can be used to computationally obtain MSE optimal variant of FISTA for performing statistical estimation. International Traveling Workshop on Interactions Between Sparse Models and Technology (iTWIST) This work may not be copied or reproduced in whole or in part for any commercial purpose. Permission to copy in whole or in part without payment of fee is granted for nonprofit educational and research purposes provided that all such whole or partial copies include the following: a notice that such copying is by permission of Mitsubishi Electric Research Laboratories, Inc.; an acknowledgment of the authors and individual contributions to the work; and all applicable portions of the copyright notice. Copying, reproduction, or republishing for any other purpose shall require a license with payment of fee to Mitsubishi Electric Research Laboratories, Inc. All rights reserved. Copyright c © Mitsubishi Electric Research Laboratories, Inc., 2016 201 Broadway, Cambridge, Massachusetts 02139 Learning MMSE Optimal Thresholds for FISTA Ulugbek S. Kamilov and Hassan Mansour. Mitsubishi Electric Research Laboratories (MERL) 201 Broadway, Cambridge, MA 02139, USA email: {kamilov, mansour}@merl.com. Abstract— Fast iterative shrinkage/thresholding algorithm (FISTA) is one of the most commonly used methods for solving linear inverse problems. In this work, we present a scheme that enables learning of optimal thresholding functions for FISTA from a set of training data. In particular, by relating iterations of FISTA to a deep neural network (DNN), we use the error backpropagation algorithm to find thresholding functions that minimize mean squared error (MSE) of the reconstruction for a given statistical distribution of data. Accordingly, the scheme can be used to computationally obtain MSE optimal variant of FISTA for performing statistical estimation. Fast iterative shrinkage/thresholding algorithm (FISTA) is one of the most commonly used methods for solving linear inverse problems. In this work, we present a scheme that enables learning of optimal thresholding functions for FISTA from a set of training data. In particular, by relating iterations of FISTA to a deep neural network (DNN), we use the error backpropagation algorithm to find thresholding functions that minimize mean squared error (MSE) of the reconstruction for a given statistical distribution of data. Accordingly, the scheme can be used to computationally obtain MSE optimal variant of FISTA for performing statistical estimation.