Informed Non-Convex Robust Principal Component Analysis With Features
Authors: Niannan Xue, Jiankang Deng, Yannis Panagakis, Stefanos Zafeiriou
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Aptly designed synthetic experiments demonstrate that our method is the first to wholly harness the power of nonconvexity over convexity in terms of both recoverability and speed. ... Extensive experimental results on synthetic data indicate that the proposed algorithm is faster and more accurate in low-rank matrix recovery than the compared state-of-the-art convex and non-convex methods for RPCA (with and without features). Experiments on two real-world datasets, namely MNIST and Yale B database demonstrate the practical merits of the proposed algorithm. |
| Researcher Affiliation | Academia | Niannan Xue,1 Jiankang Deng,1 Yannis Panagakis,12 Stefanos Zafeiriou13 1Department of Computing, Imperial College London, UK 2Department of Computer Science, Middlesex University, UK 3 Center for Machine Vision and Signal Analysis, University of Oulu, Finland |
| Pseudocode | Yes | Algorithm 1 Non-convex solver for robust principal component analysis with features |
| Open Source Code | No | The paper does not provide an explicit statement about the release of its source code or a link to a code repository for the methodology described. |
| Open Datasets | Yes | The MNIST dataset is such an example which contains hand-written digits divided into training and testing sets. The extended Yale B database is used as our observation which consists images under different illuminations for a fixed pose. |
| Dataset Splits | No | The paper mentions that the MNIST dataset is 'divided into training and testing sets' but does not provide specific percentages, sample counts, or details for these splits, nor does it explicitly mention a validation set. |
| Hardware Specification | No | The paper discusses running times and experimental comparisons but does not provide specific hardware details such as GPU/CPU models, processor types, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions algorithms and tools like SVM but does not provide specific software dependencies with version numbers (e.g., library versions, programming language versions) required for replication. |
| Experiment Setup | Yes | We have found that when the step size is set to 0.5, reasonable results can be obtained. For all algorithms in comparison, we run a total of 3000 iterations or until M L S F / M F < 10 7 is met. For PCPF, we take d = 300... and for IRPCA-IHT and our algorithm we use d = 150 instead. For fast RPCA and our algorithm, a sparsity of 0.2 is adopted. We set c to 40, t to 40 and used 10 iterations. |