Multi-class SVMs: From Tighter Data-Dependent Generalization Bounds to Novel Algorithms

Authors: Yunwen Lei, Urun Dogan, Alexander Binder, Marius Kloft

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Benchmarks on several real-world datasets show that the proposed algorithm can achieve significant accuracy gains over the state of the art.
Researcher Affiliation Collaboration Yunwen Lei Department of Mathematics City University of Hong Kong yunwelei@cityu.edu.hk Ur un Dogan Microsoft Research Cambridge CB1 2FB, UK udogan@microsoft.com Alexander Binder ISTD Pillar Singapore University of Technology and Design Machine Learning Group, TU Berlin alexander binder@sutd.edu.sg Marius Kloft Department of Computer Science Humboldt University of Berlin kloft@hu-berlin.de
Pseudocode Yes Algorithm 1: Training algorithm for ℓp-norm MC-SVM.
Open Source Code No The paper states 'We implemented the proposed ℓp-norm MC-SVM algorithm (Algorithm 1) in C++' but does not provide any link or explicit statement about releasing the source code.
Open Datasets Yes We experiment on six benchmark datasets: the Sector dataset studied in [26], the News 20 dataset collected by [27], the Rcv1 dataset collected by [28], the Birds 15, Birds 50 as a part from [29] and the Caltech 256 collected by griffin2007caltech.
Dataset Splits Yes We employ a 5-fold cross validation on the training set to tune the regularization parameter C by grid search over the set {2 12, 2 11, . . . , 212} and p from 1.1 to 2 with 10 equidistant points.
Hardware Specification No The paper does not specify any hardware components (e.g., CPU, GPU models) used for running the experiments.
Software Dependencies No The paper mentions 'implemented the proposed ℓp-norm MC-SVM algorithm (Algorithm 1) in C++' but does not list specific software dependencies with version numbers.
Experiment Setup Yes We employ a 5-fold cross validation on the training set to tune the regularization parameter C by grid search over the set {2 12, 2 11, . . . , 212} and p from 1.1 to 2 with 10 equidistant points.