Boolean Matrix Factorization and Noisy Completion via Message Passing
Authors: Siamak Ravanbakhsh, Barnabas Poczos, Russell Greiner
ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our empirical study demonstrates that message passing is able to recover low-rank Boolean matrices, in the boundaries of theoretically possible recovery and compares favorably with state-of-the-art in real-world applications, such collaborative filtering with large-scale Boolean data. 5. Experiments We evaluated the performance of message passing on random matrices and real-world data. |
| Researcher Affiliation | Academia | Siamak Ravanbakhsh MRAVANBA@CS.CMU.EDU Barnabas PoczOs BAPOCZOS@CS.CMU.EDU Carnegie Mellon University, 5000 Forbes Ave., Pittsburgh, PA 15217 USA Russell Greiner RGREINER@UALBERTA.CA University of Alberta, Edmonton, AB T6G 2E8, Canada |
| Pseudocode | Yes | Algorithm 1: message passing for Boolean matrix factorization/completion |
| Open Source Code | Yes | The Python implementation is available at https://github.com/mravanba/Boolean Factorization |
| Open Datasets | Yes | Movie Lense-1M and Movie Lense-100K dataset (http://grouplens.org/datasets/movielens/) and The senate data was obtained from http://www.stat. columbia.edu/ jakulin/Politics/senate-data.zip prepared by Jakulin et al. (2009). |
| Dataset Splits | No | No. The paper describes using a 'random subset' or 'α (0, 1) portion' of data for observation and the 'remaining (1 α portion)' for prediction, which implies a train-test split, but it does not explicitly detail a separate validation set split or a specific cross-validation setup for reproducibility. |
| Hardware Specification | No | No. The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, or memory specifications) used to run the experiments. |
| Software Dependencies | No | No. The paper mentions 'The Python implementation is available' and refers to software like 'NIMFA' and 'GLRM', but it does not specify version numbers for any programming languages, libraries, or other software dependencies. |
| Experiment Setup | Yes | In all experiments, message passing uses damping with λ = .4, T = 200 iterations and uniform priors p X m,k(1) = p Y k,n(1) = .5. and For NIMFA we use the default parameters of λh = λw = 1.1 and initialize the matrices using SVD. |