Reviews: Nonparametric Online Regression while Learning the Metric

This paper describes a novel algorithm for online nonparametric regression problem. It employs Mahalanobis metric to obtain a better distance measurement in the traditional online nonparametric regression framework. In terms of theoretical analysis, the proposed algorithm improves the regret bound and achieves a competitive result to the state-of-the-art. The theoretical proof is well organized and correct to the reviewer. However, the novelty of the proposed algorithm may be limited.