Multi-Class Optimal Margin Distribution Machine

Authors: Teng Zhang, Zhi-Hua Zhou

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical study further shows that mc ODM always outperforms all four versions of multi-class SVMs on all experimental data sets.
Researcher Affiliation Academia National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, China.
Pseudocode Yes Algorithm 1 Kenrel mc ODM ... Algorithm 2 Solving the sub-problem
Open Source Code No The paper does not provide any statement or link indicating that the source code for their method is openly available.
Open Datasets Yes Table 1 summarizes the statistics of these data sets. The data set size ranges from 150 to more than 581,012, and the dimensionality ranges from 4 to more than 62,061. Moreover, the number of class ranges from 3 to 1,000, so these data sets cover a broad range of properties.
Dataset Splits Yes For all the methods, the regularization parameter λ for mc ODM or C for binary SVM and mc SVM is selected by 5-fold cross validation from [20, 22, . . . , 220].
Hardware Specification Yes All the experiments are performed with MATLAB 2012b on a machine with 8 2.60 GHz CPUs and 32GB main memory.
Software Dependencies Yes All the experiments are performed with MATLAB 2012b on a machine with 8 2.60 GHz CPUs and 32GB main memory. The binary SVM used in ova SVM, ovo SVM and mc SVM are both implemented by the LIBLINEAR (Fan et al., 2008) package.
Experiment Setup Yes For all the methods, the regularization parameter λ for mc ODM or C for binary SVM and mc SVM is selected by 5-fold cross validation from [20, 22, . . . , 220]. For mc ODM, the regularization parameters µ and θ are both selected from [0.2, 0.4, 0.6, 0.8].