Solving Most Systems of Random Quadratic Equations

Authors: Gang Wang, Georgios Giannakis, Yousef Saad, Jie Chen

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive numerical tests using both synthetic data and real images corroborate its improved signal recovery performance and computational efficiency relative to state-of-the-art approaches.
Researcher Affiliation Academia Key Lab of Intell. Contr. and Decision of Complex Syst., Beijing Inst. of Technology Digital Tech. Center & Dept. of Electrical and Computer Eng., Univ. of Minnesota Department of Computer Science and Engineering, Univ. of Minnesota {gangwang, georgios, saad}@umn.edu; chenjie@bit.edu.cn.
Pseudocode Yes Algorithm 1 Reweighted Amplitude Flow
Open Source Code Yes For reproducibility, the Matlab code of the RAF algorithm is publicly available at https://gangwg.github.io/RAF/.
Open Datasets No The paper uses 'synthetic data' where the true signal vector x was randomly generated using x N(0, I), and the i.i.d. sensing vectors ai ai N(0, I). It also mentions a 'Galaxy image' downloaded from 'http://pics-about-space.com/milky-way-galaxy.', but this is an image, not a structured public dataset with access details.
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning.
Hardware Specification Yes All experiments were performed using MATLAB on an Intel CPU @ 3.4 GHz (32 GB RAM) computer.
Software Dependencies No The paper mentions 'MATLAB' as the software used but does not provide a specific version number or other software dependencies with version numbers.
Experiment Setup Yes Algorithm 1: 'maximum number of iterations T; step size µt = 2/6 and weighting parameter βi = 10/5 for real/complex Gaussian model; |S| = 3m/13 , and γ = 0.5.' Section 2.3: 'we take |S| := 3m/13 , βi β := 10, γ := 0.5, and µ := 2.' Section 4: 'Each scheme obtained the initial guess based on 200 power iterations, followed by a series of T = 2, 000 (truncated/reweighted) gradient iterations.'