Compressed Factorization: Fast and Accurate Low-Rank Factorization of Compressively-Sensed Data
Authors: Vatsal Sharan, Kai Sheng Tai, Peter Bailis, Gregory Valiant
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We support these theoretical results with experiments on synthetic data and demonstrate the practical applicability of compressed factorization on real-world gene expression and EEG time series datasets. |
| Researcher Affiliation | Academia | 1Stanford University, USA. |
| Pseudocode | Yes | Algorithm 1 Compressed Matrix Factorization Algorithm 2 Compressed CP Tensor Decomposition |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository. |
| Open Datasets | Yes | We evaluated our proposed NMF approach on gene expression datasets targeting three disease classes: embryonal central nervous system tumors (Pomeroy et al., 2002), lung carcinomas (Bhattacharjee et al., 2001), and leukemia (Mills et al., 2009) (Table 1)... We experimented with tensor decomposition on a compressed tensor derived from the CHB-MIT Scalp EEG Database (Shoeb & Guttag, 2010). |
| Dataset Splits | No | The paper describes using synthetic and real-world datasets, and evaluates 'approximation error' and 'reconstruction error' for the entire dataset or its components. It does not specify explicit train/validation/test dataset splits with percentages or sample counts. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to run the experiments. |
| Software Dependencies | No | For sparse PCA, we use alternating minimization with LARS (Zou et al., 2006), and for NMF, we use projected gradient descent (Lin, 2007). These are general methods or algorithms, but the paper does not specify any software names with version numbers for implementation (e.g., Python, PyTorch, scikit-learn versions). |
| Experiment Setup | Yes | For all datasets, we fixed rank r = 10... We ran projected gradient descent (Lin, 2007) for 250 iterations, which was sufficient for convergence... For sparse PCA, we report results for the setting of the ℓ1 regularization parameter that yielded the lowest approximation error. We did not use an ℓ1 penalty for NMF. We give additional details in the Appendix. |