Improving Efficiency of SVMk-Fold Cross-Validation by Alpha Seeding

Authors: Zeyi Wen, Bin Li, Ramamohanarao Kotagiri, Jian Chen, Yawen Chen, Rui Zhang

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results show that our algorithms are several times faster than the k-fold cross-validation which does not make use of the previously trained SVM. Moreover, our algorithms produce the same results (hence same accuracy) as the k-fold cross-validation which does not make use of the previously trained SVM.
Researcher Affiliation Academia 1{zeyi.wen, kotagiri, rui.zhang}@unimelb.edu.au The University of Melbourne, Australia 2{gitlinux@gmail.com, ellachen@scut.edu.cn, elfairyhyuk@gmail.com} South China University of Technology, China
Pseudocode No The pseudo-code of the full algorithm is shown in Algorithm 1 in Wen et al. (2016).
Open Source Code No The paper does not provide an explicit statement or link for open-source code related to the methodology described.
Open Datasets Yes We empirically evaluate our proposed algorithms using five datasets from the Lib SVM website (Chang and Lin 2011).
Dataset Splits Yes The k-fold cross-validation is a commonly used process to evaluate the effectiveness of SVMs with the selected hyper-parameters. We varied k from 3 to 100 to study the effect of the value of k.
Hardware Specification Yes The experiments were conducted on a desktop computer running Linux with a 6-core E5-2620 CPU and 128GB main memory.
Software Dependencies No All our proposed algorithms were implemented in C++. We empirically evaluate our proposed algorithms using five datasets from the Lib SVM website (Chang and Lin 2011). The paper mentions software used (C++, Lib SVM) but does not provide specific version numbers for software dependencies.
Experiment Setup Yes Following the common settings, we used the Gaussian kernel function and by default k is set to 10. The hyper-parameters for each dataset are identical to the existing studies (Catanzaro, Sundaram, and Keutzer 2008; Smirnov, Sprinkhuizen-Kuyper, and Nalbantov 2004; Wu and Li 2006). Table 2 gives more details about the datasets.