Scaling Properties of Deep Residual Networks
Alain–Sam Cohen 1 Rama Cont 2 Alain Rossier 2 1 Renyuan Xu 2
(L) (L)
Abstract where hk is the hidden state at layer k = 0, . . . , L, h0 =
(L)
Residual networks (ResNets) have displayed im- x ∈ Rd the input, hL ∈ Rd the output, σ : R → R is a non-
pressive results in pattern recognition and, re- ...


雷达卡




京公网安备 11010802022788号







