githublzb / GLISTA

source code of GLISTA

GLISTA

source code of GLISTA

Sparse Coding with Gated Learned ISTA

This repository contains an implementation of the simulation experiments in Sparse Coding with Gated Learned ISTA. The code base is based from the LISTA-CPSS repo (https://github.com/xchen-tamu/linear-lista-cpss). You should add this file into the "model" in LISTA-CPSS repo

Dependencies

• tensorflow

Usage

The code supports GLISTA, LISTA, LAMP, LISTA-SS, and LISTA-CP-SS. To run our experiment, use "main.py". For GLISTA with a combination gain gate functions with coupled parameters under SNR = 10dB, run:

python main.py --net GLISTA_cp \ 
			   --SNR 10 \
			   --gpu 0 \
			   -M 250 \
			   -N 500 \
			   -gain \
			   -fixval False \
			   -u \
			   -a 5.0 \
			   -uf combine

For GLISTA with an inverse overshoot gate functions under SNR = 40dB, run:

python main.py --net GLISTA \ 
			   --SNR 40 \
			   --gpu 0 \
			   -M 250 \
			   -N 500 \
			   -o \
			   -fixval False \
			   -a 1.0 \
			   -uf inv \