Newton Method over Networks is Fast up to the Statistical Precision
Amir Daneshmand 1 Gesualdo Scutari 1 Pavel Dvurechensky 2 3 Alexander Gasnikov 4 3
Abstract where ` : Rd × Z → R is the loss function, assumed to be
(strongly) convex in x, for any given z ∈ Z. Agents aim to
We propose a distributed cubic regularization of minimize the total empirical risk over the N = mn samples,
the Newton method for sol ...


雷达卡




京公网安备 11010802022788号







