Fast Stochastic Alternating Direction Method of Multipliers
Authors: Wenliang Zhong, James Kwok
ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on the graph-guided fused lasso demonstrate that the new algorithm is significantly faster than state-of-the-art stochastic and batch ADMM algorithms. |
| Researcher Affiliation | Academia | Leon Wenliang Zhong WZHONG@CSE.UST.HK James T. Kwok JAMESK@CSE.UST.HK Department of Computer Science and Engineering, Hong Kong University of Science and Technology, Hong Kong |
| Pseudocode | Yes | Algorithm 1 Stochastic average alternating direction method of multipliers (SA-ADMM). |
| Open Source Code | No | The paper does not provide any concrete access information (e.g., repository link, explicit statement of code release, or mention of code in supplementary materials) for the source code of the described methodology. |
| Open Datasets | Yes | Experiments are performed on four popular binary classification data sets2 (Table 2) (Le Roux et al., 2012; Suzuki, 2013). a9a, covertype and rcv1 are from the LIBSVM archive, and sido from the Causality Workbench website. |
| Dataset Splits | Yes | For each data set, half of the samples are used for training, while the rest for testing. To reduce statistical variability, results are averaged over 10 repetitions. [...] For ρ in (3) and the stepsize (or its proportionality constant), we use a small training subset with 500 samples, and choose the parameter setting with the smallest training objective value after running the stochastic algorithm over 5 data passes; or after running the batch algorithm for 100 iterations. |
| Hardware Specification | Yes | All methods are implemented in MATLAB, and experiments are performed on a PC with an Intel i7-2600K CPU and 32GB memory. |
| Software Dependencies | No | All methods are implemented in MATLAB. However, no specific version number for MATLAB or any other software dependencies are provided. |
| Experiment Setup | Yes | we fix the regularization parameter λ to 10 5 for a9a, covertype, and to 10 4 for rcv1 and sido. For ρ in (3) and the stepsize (or its proportionality constant), we use a small training subset with 500 samples, and choose the parameter setting with the smallest training objective value after running the stochastic algorithm over 5 data passes; or after running the batch algorithm for 100 iterations. |