Towards Multi-Mode Outlier Robust Tensor Ring Decomposition

Authors: Yuning Qiu, Guoxu Zhou, Andong Wang, Zhenhao Huang, Qibin Zhao

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Experimental Results In this section, we evaluate the performance of the proposed approach by conducting experiments on both synthetic data and real-world datasets, including light field images and hyperspectral videos. We compare the experimental results with some state-of-the-art robust matrix/tensor decomposition approaches...
Researcher Affiliation Academia 1 School of Automation, Guangdong University of Technology, Guangzhou, 510006, China 2 RIKEN Center for Advanced Intelligence Project, Tokyo, 1030027, Japan 3 Key Laboratory of Intelligent Detection and The Internet of Things in Manufacturing, Ministry of Education, Guangzhou, 510006, China
Pseudocode No The optimization algorithm employed to solve the Eq. (5) hinges on the utilization of the Alternating Direction Method of Multipliers (ADMM) algorithm (Boyd et al. 2011). For a comprehensive understanding, please refer to Appendix B.
Open Source Code Yes The implementation code is available at https://github.com/ynqiu/MORTRD.
Open Datasets Yes We randomly select four HSV datasets2. ... https://www.hsitracking.com/contest/ We adopt a publicly accessible light field images dataset3, and randomly select 6 of these images. ... https://lightfield-analysis.uni-konstanz.de/
Dataset Splits No No explicit specification of training, validation, and test dataset splits (e.g., percentages, sample counts, or cross-validation setup) was found in the main text.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, processor types, memory amounts) used for running experiments were provided in the paper.
Software Dependencies No The paper mentions the use of the Alternating Direction Method of Multipliers (ADMM) algorithm but does not specify any software libraries or dependencies with version numbers.
Experiment Setup Yes To generate synthetic low rank tensor T P Rd1ˆd2ˆ ˆd K with TR rank rr1, r2, , r Ks, we first generated K core tensors Gpkq P Rrkˆdkˆrk 1 where each entry is produced by the i.i.d. Gaussian distribution Np0, 1q. To construct the latent structural tensor Sk, , we let the support set of Sk pkq be Ωk, and then randomly select |Ωk pkq| columns of Sk pkq as outliers whose entries obey i.i.d. Np0, 1q. Thus, the outlier is given by S řK k 1 Sk. The additive noise tensor is produced by Np0, σ2q, where σ 10 3}T }F{ ? D to guarantee a constant signal-to-noise ratio (SNR). All the experiments are repeated 10 times and their mean values are reported. ... The additive Gaussian noise is set as Section 5.1 with σ 0.05}T }F{ ? D. The multi-mode outliers are generated with |Ωk pkq| roundp10 2 ś j,j k djq, and each outlier Sk pkqp:, iq is generated by a uniform discrete distribution on r 1, 1s.