Learning Sparse Confidence-Weighted Classifier on Very High Dimensional Data

Authors: Mingkui Tan, Yan Yan, Li Wang, Anton Van Den Hengel, Ivor W. Tsang, Qinfeng (Javen) Shi

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experimental results demonstrate the superior performance of the proposed methods over state-of-the-art counterparts on classification and feature selection tasks. Experiments We conduct two sets of experiments to verify our methods.
Researcher Affiliation Academia ACVT, The University of Adelaide, Australia QCIS, University of Technology Sydney, Australia Department of Mathematics, Statistics, and Computer Science, University of Illinois at Chicago, USA {mingkui.tan, anton.vandenhengel, javen.shi}@adelaide.edu.au, {yan.yan,ivor.tsang}@uts.edu.au, liwang8@uic.edu
Pseudocode Yes Algorithm 1: Learning confidence-weighted classifiers in online-batch. Algorithm 2: Sparse CW learning in online-batch. Algorithm 3: Online proximal primal-dual coordinate ascent for solving problem (13)
Open Source Code Yes The sources of our methods are available from http://www.tanmingkui.com/sbcw.html.
Open Datasets Yes All data sets are widely used benchmarks in machine learning, and are summarized in Table 1. URL is originally from http://sysnet.ucsd.edu/projects/url/ for identifying suspicious URLs, astro-ph is from (Hsieh et al. 2008), and others are from http://www.csie.ntu.edu.tw/ cjlin/libsvmtools/datasets/.
Dataset Splits Yes For all competing methods, we apply 5-cross-validation to choose their parameters.
Hardware Specification Yes All experiments are conducted on a PC installed a 64-bit operating system with an Intel(R) Core(TM) Xeon CPU 3.00GHz and 32GB memory.
Software Dependencies No The paper mentions software like 'Liblinear package' and 'libol.stevenhoi.org' but does not provide specific version numbers for these software dependencies, only general links or names.
Experiment Setup Yes Require: Parameters r, C > 0, and μ = 0 and Σ = I. (Algorithm 1) in our experiments on feature selection task, we stop SBCW after miter = 15 iterations, and set r = p/miter in order to select p features