Linear Spectral Estimators and an Application to Phase Retrieval

Authors: Ramina Ghods, Andrew Lan, Tom Goldstein, Christoph Studer

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We now compare the performance of our LSPEs against existing spectral initializers proposed for phase retrieval on synthetic and real image data. All our results use the spectral initializers and experimental setups provided by Phase Pack (Chandra et al., 2017).
Researcher Affiliation Academia 1School of Electrical and Computer Engineering, Cornell University, Ithaca, NY 2Department of EE, Princeton University 3University of Maryland. Correspondence to: Ramina Ghods <rg548@cornell.edu>, Christoph Studer <studer@cornell.edu>.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper mentions using 'Phase Pack (Chandra et al., 2017)' but does not provide concrete access to the source code for their own proposed methodology (LSPEs).
Open Datasets Yes Our goal is to recover a 16 16-pixel and a 40 40-pixel image that was captured through a multiple scattering media using the deterministic and highly-structured transmission matrix as detailed in (Metzler et al., 2017).
Dataset Splits No The paper defines measurement parameters like M and N and problem setups, but does not provide specific training/validation/test dataset splits in terms of percentages, sample counts, or citations to predefined splits for a machine learning model.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No All our results use the spectral initializers and experimental setups provided by Phase Pack (Chandra et al., 2017). The paper mentions 'Phase Pack' but does not provide a version number for this or any other software dependency.
Experiment Setup Yes In the following experiment, we set M = 8N and vary the dimension N from 8 to 64. For each pair (M, N), we randomly generate one instance of an i.i.d. circularly symmetric complex Gaussian measurement matrix and average the different errors (S-MSE and EER) over 10, 000 Monte-Carlo trials. We consider a noiseless setting and assume identity preprocessing, i.e., T (y) = y. The signal vectors are generated according to an i.i.d. circularly complex Gaussian random vector.