Efficient Compressive Phase Retrieval with Constrained Sensing Vectors

Authors: Sohail Bahmani, Justin Romberg

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also evaluate the algorithm through numerical simulation.We evaluated the performance of Algorithm 1 through some numerical simulations.
Researcher Affiliation Academia Sohail Bahmani, Justin Romberg School of Electrical and Computer Engineering. Georgia Institute of Technology Atlanta, GA 30332 {sohail.bahmani,jrom}@ece.gatech.edu
Pseudocode Yes Algorithm 1: input : the measurements y, the operator W, and the matrix Ψ output: the estimate c X 1 Low-rank estimation stage: b B argmin B 0 trace (B) subject to W (B) y 2 ε (6) 2 Sparse estimation stage: c X argmin X X 1 subject to ΨXΨ T b B F Cε n
Open Source Code No The paper mentions using the TFOCS package for implementation but does not state that the authors are releasing their own code for the methodology described.
Open Datasets No We considered the target k-sparse signal x to be in R256 (i.e., d = 256). The support set of of the target signal is selected uniformly at random and the entry values on this support are drawn independently from N (0, 1). The noise vector z is also Gaussian with independent N 0, 10 4 . The operator W and the matrix Ψ are drawn from some Gaussian ensembles as described in Corollary 1. The paper describes how it generates synthetic data for its experiments, but it does not refer to or provide access information (link, DOI, citation) for a pre-existing publicly available or open dataset.
Dataset Splits No The paper describes generating synthetic data for numerical simulations but does not specify training, validation, or test dataset splits. It mentions '100 trials' and varying 'k' and 'm,n' for experiment setup but not data partitioning.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running the experiments. It only mentions 'numerical simulations'.
Software Dependencies No The low-rank estimation stage and the sparse estimation stage are implemented using the TFOCS package [24]. While a package is named, no specific version number for TFOCS or any other software dependency is provided.
Experiment Setup Yes We considered the target k-sparse signal x to be in R256 (i.e., d = 256). The support set of of the target signal is selected uniformly at random and the entry values on this support are drawn independently from N (0, 1). The noise vector z is also Gaussian with independent N 0, 10 4 . The operator W and the matrix Ψ are drawn from some Gaussian ensembles as described in Corollary 1. In the first experiment, for each value of k, the pair (m, n) that determines the size W and Ψ are selected from {(8k, 24k) , (8k, 32k) , (12k, 36k) , (12k, 48k) , (16k, 48k)}. In the second experiment we compared the performance of Algorithm 1 to the convex optimization methods that do not exploit the structure of the sensing vectors. The setup for this experiment is the same as in the first experiment except for the size of W and Ψ; we chose m = 2k 1 + log d k and n = 3m, where r denotes the smallest integer greater than r.