Linear Probability Forecasting - Computer Science > LearningReport as inadecuate

Linear Probability Forecasting - Computer Science > Learning - Download this document for free, or read online. Document in PDF available to download.

Abstract: Multi-class classification is one of the most important tasks in machinelearning. In this paper we consider two online multi-class classificationproblems: classification by a linear model and by a kernelized model. Thequality of predictions is measured by the Brier loss function. We suggest twocomputationally efficient algorithms to work with these problems and provetheoretical guarantees on their losses. We kernelize one of the algorithms andprove theoretical guarantees on its loss. We perform experiments and compareour algorithms with logistic regression.

Author: Fedor Zhdanov, Yuri Kalnishkan



Related documents