Ada-LISTA: Learned Solvers Adaptive to Varying Models
This addresses the limitation of fixed-model learned solvers for sparse coding applications, offering an adaptive solution for scenarios with varying dictionaries, though it is incremental over existing LISTA methods.
The paper tackled the problem of learned iterative solvers being inapplicable to varying models by introducing Ada-LISTA, a universal architecture that adapts to different dictionaries, proving it solves sparse coding linearly and demonstrating practical adaptation in image inpainting.
Neural networks that are based on unfolding of an iterative solver, such as LISTA (learned iterative soft threshold algorithm), are widely used due to their accelerated performance. Nevertheless, as opposed to non-learned solvers, these networks are trained on a certain dictionary, and therefore they are inapplicable for varying model scenarios. This work introduces an adaptive learned solver, termed Ada-LISTA, which receives pairs of signals and their corresponding dictionaries as inputs, and learns a universal architecture to serve them all. We prove that this scheme is guaranteed to solve sparse coding in linear rate for varying models, including dictionary perturbations and permutations. We also provide an extensive numerical study demonstrating its practical adaptation capabilities. Finally, we deploy Ada-LISTA to natural image inpainting, where the patch-masks vary spatially, thus requiring such an adaptation.