A single pairwise model for classification using online learning with kernels
Engin Taş · Hacettepe Journal of Mathematics and Statistics · 2017
Any binary or multi-class classi cation problem can be transformed into a pairwise prediction problem. This expands the data and brings an advantage of learning from a richer set of examples, in the expense of increasing costs when the data is in higher dimensions. Therefore, this study proposes to adopt an online support vector machine to work with pairs of examples. This modi fied algorithm is suitable for large data sets due to its online nature and it can also handle the sparsity structure existing in the data. Performances of the pairwise setting and the direct setting are compared in two problems from different domains. Results indicate that the pairwise setting outperforms the direct setting signifi cantly. Furthermore, a general framework is designed to use this pairwise approach in a multi-class classi cation task. Result indicate that this single pairwise model achieved competitive classi cation rates even in large-scaled datasets with higher dimensionality.