Coherence Pursuit: Fast, Simple, and Robust Subspace Recovery

Authors: Mostafa Rahmani, George Atia

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5. Numerical Simulations In this section, the performance of the proposed method is investigated with both synthetic and real data. We compare the performance of Co P with the state-of-the-art robust PCA algorithms including FMS (Lerman & Maunu, 2014), GMS (Zhang & Lerman, 2014), R1-PCA (Ding et al., 2006), OP (Xu et al., 2010b), and SPCA (Maronna et al., 2006)., Table 1. Running time of the algorithms m = n Co P FMS OP R1-PCA 1000 0.02 1 45 1.45 2000 0.7 5.6 133 10.3 5000 5.6 60 811 83.3 10000 27 401 3547 598
Researcher Affiliation Academia 1University of Central Florida, Orlando, Florida, USA. Correspondence to: Mostafa Rahmani <mostafa@knights.ucf.edu>.
Pseudocode Yes Algorithm 1 Co P: Proposed Robust PCA Algorithm, Algorithm 2 Adaptive Column Sampling for the Subspace Identification Step (step 3) of Co P, Algorithm 3 Subspace Clustering Error Correction Method
Open Source Code No The paper does not contain any explicit statement about releasing open-source code for the described methodology, nor does it provide a link to a code repository.
Open Datasets Yes We use the Hopkins155 dataset which contains video sequences of 2 or 3 motions (Tron & Vidal, 2007).
Dataset Splits No The paper describes parameters for synthetic data generation (e.g., 'm = 100, r = 10', 'n1 = n/5 and n2 = 4n/5') and uses the Hopkins155 dataset, but it does not specify explicit train/validation/test dataset splits (e.g., percentages, sample counts, or predefined splits for validation) needed for reproduction.
Hardware Specification No The paper provides running times in Table 1 but does not specify any concrete hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes Algorithm 1 Co P: Proposed Robust PCA Algorithm Initialization: Set p = 1 or p = 2., For instance, consider a setting in which m = 400, r = 5, n1 = 50, and the distributions of inliers and outliers follow Assumption 1. Fig. 1 shows the vector p (c.f. Algorithm 1) for different values of p and n2., In this simulation, we construct the matrix Y using 20 columns of X corresponding to the largest 20 elements of p., In this simulation, we use 30 columns to form the matrix Y., When Co P is applied, 50 percent of the columns of X are used to form the matrix Y.