Fast and Provable Nonconvex Tensor RPCA

Authors: Haiquan Qiu, Yao Wang, Shaojie Tang, Deyu Meng, Quanming Yao

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental These benefits are also empirically verified both on synthetic data, and real applications, e.g., hyperspectral image denoising and video background subtraction.
Researcher Affiliation Collaboration 1 Xi an Jiaotong University, Xi an, China 2 Tsinghua University, Beijing, China 3 4Paradigm Inc., Beijing, China 4 The University of Texas at Dallas, Texas, USA 5 Macau University of Science and Technology, Macau, China.
Pseudocode Yes Algorithm 1 Initialization; Algorithm 2 APT: Alternating Projection Algorithm for Tensor RPCA; Algorithm 3 Trim for tensor; Algorithm 4 EAPT: Efficient Alternating Projection Algorithm for Tensor RPCA
Open Source Code No The paper does not provide any statement or link indicating the public availability of source code for the described methodology.
Open Datasets Yes We use the CAVE dataset (Yasuma et al., 2008) for experiments.
Dataset Splits No The paper mentions selecting images and frames for experiments but does not specify train/validation/test splits, percentages, or absolute sample counts for each split.
Hardware Specification Yes All experiments are conducted on a PC with two Intel Xeon Silver 4215R 3.20GHz CPUs and 187GB memory with MATLAB R2021b.
Software Dependencies Yes All experiments are conducted on a PC with two Intel Xeon Silver 4215R 3.20GHz CPUs and 187GB memory with MATLAB R2021b.
Experiment Setup Yes We set n = 500, q = 5, r = 5, α = 0.3 in Figure 1. The hyperparameters of our methods are set as the same suggested by Proposition 5.3, Theorems 5.4 and 5.5. The hyperparameters of other methods are listed in the Appendix C.1.