Linear algorithms for online multitask classification

We introduce new Perceptron-based algorithms for the online multitask binary classification problem. Under suitable regularity conditions, our algorithms are shown to improve on their baselines by a factor proportional to the number of tasks. We achieve these improvements using various types of regularization that bias our algorithms towards specific notions of task relatedness. More specifically, similarity among tasks is either measured in terms of the geometric closeness of the task reference vectors or as a function of the dimension of their spanned subspace. In addition to adapting to the online setting a mix of known techniques, such as the multitask kernels of Evgeniou et al., our analysis also introduces a matrix-based multitask extension of the p-norm Perceptron, which is used to implement spectral co-regularization. Experiments on real-world data sets complement and support our theoretical findings.