Fast Compressive Phase Retrieval under Bounded Noise

Authors: Hongyang Zhang, Shan You, Zhouchen Lin, Chao Xu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on synthetic and real data verify our theory. We implement numerical simulations to evaluate the performance of the proposed Algorithm 1 and compare it with other competing methods. Synthetic Simulations and Real Experiment: Image Recovery.
Researcher Affiliation Academia 1Machine Learning Department, Carnegie Mellon University, U.S.A. 2Key Lab. of Machine Perception (MOE), School of EECS, Peking University, P. R. China 3 Cooperative Medianet Innovation Center, Shanghai Jiao Tong University, P. R. China
Pseudocode Yes Algorithm 1 Robust Recovery of Sparse Signal by Phase Retrieval
Open Source Code No The paper states 'for the Recover step in Algorithm 1 we utilize the SPAMS toolbox (Mairal et al. 2009)' but does not provide any link or statement about releasing the source code for their own method.
Open Datasets No The paper describes generating synthetic data for simulations and uses a single image (from NASA) for a real-world example, but it does not specify any publicly available datasets used for training with access information (link, DOI, or formal citation).
Dataset Splits No The paper describes generating synthetic data for experiments and conducting '100 trials' to investigate recovery performance. However, it does not specify any training, validation, or test splits of a fixed dataset.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments. It only mentions 'laptop computer' in the context of memory limitations for other algorithms.
Software Dependencies No The paper states 'for the Recover step in Algorithm 1 we utilize the SPAMS toolbox (Mairal et al. 2009)' but does not provide a specific version number for this or any other software dependency.
Experiment Setup Yes We fix the dimension of the groundtruth sparse real signal x0 to be 256 (i.e. n = 256). We select the support size t in set {2, 4, 6, ..., 20} and vary the values of d and m to investigate the influence of their choices on the recovery accuracy of our method. All experiments are implemented over 100 trials and for the Recover step in Algorithm 1 we utilize the SPAMS toolbox (Mairal et al. 2009). We choose d = 2t(1+log n/t) , m = 10d, σ2 = 10^-2 in different t {6, 8, 10, 12, 14}. Specifically, d1 = 2t(1 + log n/t) and d2 = 2t log n/t.