Robust Tensor Decomposition via Orientation Invariant Tubal Nuclear Norms
Authors: Andong Wang, Chao Li, Zhong Jin, Qibin Zhao6102-6109
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on real-world datasets demonstrate the superiority and effectiveness of the proposed norms. |
| Researcher Affiliation | Academia | Andong Wang,1,2,3 Chao Li,2 Zhong Jin,*1,4 Qibin Zhao*2,3 1School of Computer Science & Engineering, NJUST, 2Tensor Learning Unit, RIKEN AIP, 3School of Automatation, GDUT, 4Key Laboratory of Intelligent Perception and System for High-Dimensional Information of Ministry of Education, NJUST Corresponding authors: Zhong Jin (zhongjin@njust.edu.cn), Qibin Zhao (qibin.zhao@riken.jp) |
| Pseudocode | Yes | Then we continue solving Model I and Model II by ADMM presented in Algorithm 1 and Algorithm 2 respectively, where all the sub-problems can be solved in closed forms6. ...6Due to space limitation, the description of Algorithms 1 and 2 are shown in the supplementary material. |
| Open Source Code | Yes | Matlab implementations can be found in https://qibinzhao.github.io |
| Open Datasets | Yes | The performance comparison is carried out on the widely used seven YUV videos7: akiyo, bridge-far, carphone, claire, coastguard, container and foreman. ...7The videos are available from https://sites.google.com/site/subudhibadri/fewhelpfuldownloads. ...8Scenario B and Scenario B-additional dataset from http://www.mrt.kit.edu/z/publ/download/velodynetracking/dataset.html |
| Dataset Splits | No | The paper does not explicitly describe a validation dataset split. The experiments involve data recovery or inpainting where the goal is to recover original data from corrupted versions, and performance is evaluated on the recovered results using PSNR against ground truth. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware specifications (e.g., CPU, GPU models, memory) used to conduct the experiments. |
| Software Dependencies | No | The paper mentions 'Matlab implementations' for its algorithms but does not specify a version number for Matlab or any other specific software libraries or dependencies with their versions. |
| Experiment Setup | Yes | Parameters for OITNN-O are set as w1 : w2 : w3 = a1 : 1 : a1, with a1 [0.1, 0.5], and OITNN-L v1 : v2 : v3 = 1 : a2 : 1 with a2 [0.025, 0.055]. The weight parameters α of SNN are chosen to satisfy α1 : α2 : α3 = 1 : 1 : 0.01 as suggested in (Liu et al. 2013). The sparse/low-rank parameter ratio μ/λ of NN is 1/ d2 (Cand es et al. 2011), and 1/ 3d2 for Sq NN, TNN and t-TNN. We tune the sparse/low-rank parameter ratio for OITNN-O and OITNN-L. |