Neural Proximal Gradient Descent for Compressive Imaging

Authors: Morteza Mardani, Qingyun Sun, David Donoho, Vardan Papyan, Hatef Monajemi, Shreyas Vasanawala, John Pauly

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments are carried out under different settings: (a) reconstructing abdominal MRI of pediatric patients from highly undersampled Fourier-space data and (b) superresolving natural face images.
Researcher Affiliation Academia Depts. of 1Electrical Eng., 2Radiology, 3Statistics, and 4Mathematics; Stanford University
Pseudocode No The paper describes the iterative procedure in mathematical equations and prose but does not provide a formal pseudocode block or algorithm box.
Open Source Code Yes The source code for Tensor Flow implementation is publicly available in the Github page [35].
Open Datasets Yes Adopting celeb Faces Attributes Dataset (Celeb A) [40], for training and test we use 10K and 1, 280 images, respectively.
Dataset Splits No The paper mentions 'train dataset' and 'test dataset' with specific counts, but does not explicitly describe a validation split.
Hardware Specification Yes Training was performed with Tensor Flow interface on an NVIDIA Titan X Pascal GPU with 12GB RAM.
Software Dependencies No The paper mentions 'Tensor Flow interface' but does not specify its version number or any other software dependencies with versions.
Experiment Setup Yes We used the Adam SGD optimizer with the momentum parameter 0.9, mini-batch size 2, and initial learning rate 10 5 that is halved every 10K iterations. For training RNN, we use ℓ2 cost in (P2) with β = 0.75.