Non-negative Matrix Factorization under Heavy Noise
Authors: Chiranjib Bhattacharya, Navin Goyal, Ravindran Kannan, Jagdeep Pani
ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We provide empirical justification for our assumptions on C. Our algorithm outperforms earlier polynomial time algorithms both in time and error, particularly in the presence of high noise. In this section we do a comprehensive empirical evaluation of TSVDNMF on synthetic and real datasets. |
| Researcher Affiliation | Collaboration | Department of Computer Science and Automation, Indian Institute of Science, Bangalore, India Microsoft Research India |
| Pseudocode | Yes | TSVDNMF Input: A,k, α, ε, ε4, β + ρ, ε0 Output: Basis matrix B. 1. Thresholding: Apply the Thresholding procedure (see below) to get D. 2. SVD: Find the best rank k approximation D(k) to D. |
| Open Source Code | No | The paper mentions that 'Codes for SPA, PW-SPA, Gillis-LP are obtained from their first author s website' (footnote 9 in Section 5.2), referring to baselines used for comparison. However, it does not provide any explicit statement or link for the open-sourcing of the authors' own TSVDNMF code. |
| Open Datasets | Yes | Dataset: Reuters 4, 20 Newsgroups 5, TDT-2 6, Yale face dataset7. ... (1) Reuters10 (n = 7285, d = 12418, k = 10), (2) Reuters48 (n = 8258, d = 13647, k = 48), (3) 20 Newsgroups (n = 18846, d = 24287, k = 20), (4) TDT-2 (n = 9394, d = 20687, k = 30). (5) Yale (n = 165, d = 4096, k = 15). The footnotes provide URLs to these public datasets. |
| Dataset Splits | No | The paper describes the datasets used and the overall experimental setup, but it does not provide specific details regarding training, validation, or test data splits (e.g., percentages, sample counts, or explicit references to standard splits with citations) for reproducibility. |
| Hardware Specification | Yes | All experiments are performed using Matlab on a system with 3.5 GHz processor and 8GB RAM. |
| Software Dependencies | No | The paper states that experiments were performed using 'Matlab' but does not specify a version number for Matlab or any other software libraries or dependencies. This lack of version information makes it difficult to reproduce the software environment. |
| Experiment Setup | Yes | We fixed the other constants as ε0 = 0.04, α = 0.9, ν = 1.15 in all the experiments. |