文摘
We introduce a modification to the well studied leverage score sampling algorithm which takes into account data scale, called the augmented leverage score, and introduce an initial error bound proof for the case of deterministic sampling – which to our knowledge is the first bound for this augmented leverage score. We discuss the implications of the error bounds proof and present an empirical evaluation of the proposed augmented leverage score performance on the column subsample selection problem (CSSP) as compared to the traditional leverage score and other methods in both a deterministic and probabilistic sampling paradigm. We show that the augmentation of the leverage score improves the empirical performance on CSSP significantly for many kinds of data.