2 Replies to “A COLT paper”

  1. John,
    Can you compare/contrast this paper with its reference [4] by Cesa-Bianchi et. al.? The later paper seemed appealingly elementary and with similar goals.

    Drew

  2. I believe [4] is superceded entirely. [4] was showing that it was possible (with bad constants) to take advantage of a small number of errors. This paper shows that we can do it efficiently, using a very fast approximation that is perfect with respect to the leading constant.

Comments are closed.