Multi_task_learning

Multi-task feature learning

This is a method for learning multiple tasks simultaneously, assuming that they share a set of common latent features. It is based on regularizing the spectrum of the matrix of tasks. Regularization with the trace norm is a special case of this framework. Among the diverse applications of multi-task learning, one example is the personalized recommendation of products to consumers.

The methodology is presented in detail in the papers Multi-Task Feature Learning, Convex Multi-Task Feature Learning and A Spectral Regularization Framework for Multi-Task Structure Learning.

Note that it is possible to use the method with a nonlinear kernel instead of explicit features. It suffices to run the code on the Gram matrix after a preprocessing with a Gram-Schmidt or Cholesky decomposition (see Convex Multi-Task Feature Learning sec. 5).