Scaling SVM and Least Absolute Deviations via Exact Data Reduction

Authors: Jie Wang, Peter Wonka, Jieping Ye

ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We have evaluated DVI on both synthetic and real data sets. Experiments indicate that DVI significantly outperforms the existing state-of-the-art screening rules for SVM, and it is very effective in discarding non-support vectors for LAD.
Researcher Affiliation Academia Jie Wang JIE.WANG.USTC@ASU.EDU Arizona State University, Tempe, AZ 85287 USA Peter Wonka PWONKA@GMAIL.COM King Abdullah University of Science and Technology, Thuwal, Saudi Arabia Arizona State University, Tempe, AZ 85287 USA Jieping Ye JIEPING.YE@ASU.EDU Arizona State University, Tempe, AZ 85287 USA
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper does not include an unambiguous statement that the authors are releasing the code for the work described in this paper, nor does it provide a direct link to a source-code repository.
Open Datasets Yes We compare the performance of DVIs, SSNSV and ESSNSV using (a) the IJCNN1 data set (Prokhorov, 2001); (b) the Wine Quality data set (Cortez et al., 2009); and (c) the Forest Covertype data set (Hettich & Bay, 1999). We evaluate the performance of DVIs for LAD on three real data sets: (a) Magic Gamma Telescope data set (Bache & Lichman, 2013); (b) Computer data set (Rasmussen et al.); (c) Houses data set (Pace & Barry, 1997).
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes We test the rules along a sequence of 100 parameters of C [10^-2, 10] equally spaced in the logarithmic scale.