Uncertainty loom for early-warning

To minimize the bound of leave-one-out error directly, a convex optimization problem can be derived which constructs a sparse linear classifier using kernel game. However, standard leave-one-out support vector machine (LOOM) can not classify patterns with uncertainty in the information input. A new LOOM is proposed which is able to deal with training data with uncertainty based on expert advices. Firstly the meaning of the uncertainty is defined. Based on this meaning of uncertainty, the algorithm has been derived. This technique extends the application horizon of LOOM greatly. As an application, the problem about early-warning of food security is solved by our algorithm.