Noisy-Sudocodes - Software
This webpage describes the
Matlab files used in our work on sparse signal reconstruction in Bayesian compressed sensing, as well as 1-bit compressed sensing, where the measurement matrix is the concatenation of a sparse binary matrix and a dense matrix with i.i.d. Gaussian entries. The construction of such a matrix allows us to design fast reconstruction algorithms; details appear in the following paper,
"Two-Part Reconstruction with Noisy-Sudocodes",
IEEE Trans. Signal Process., vol 62, no. 23, pp. 6323-6334, December 2014
The software was implemented by
The implementation is a two-part reconstruction algorithm, where partial zero-valued entries are identified in Part 1, and the remaining entries are estimated by approximate message passing (AMP), which is proposed by Donoho et al., or binary iterative hard thresholding (BIHT), which is proposed by Jacques et al., in Part 2, for Bayesian compressed sensing problems and 1-bit compressed sensing problems, respectively.
Below is a brief description of the files used in our implementation.
Yanting Ma, January 2016
Bayesian compressed sensing problems
sudo_amp.m: script that runs the Sudo-AMP algorithm, where AMP is used in Part 2.
analysisPart1.m: estimation of the sparsity level after removing the entries that are identified as zero in Part 1. The updated sparsity level will be used in Part 2.
AMP.m: implementation of the AMP algorithm for Part 2.
1-bit compressed sensing problems
sudo_BIHT1.m: script that runs the Sudo-BIHT1 algorithm, where BIHT1 is used in Part 2, for 1-bit compressed sensing problems without pre-quantization noise.
sudo_BIHT2.m: script that runs the Sudo-BIHT2 algorithm, where BIHT2 is used in Part 2, for 1-bit compressed sensing problems with pre-quantization noise.
readme.txt: instruction for downloading the BIHT.m and BIHT2.m files, which are used in sudo_BIHT1.m and sudo_BIHT2.m, respectively.
Back to my homepage.