Совместный семинар ИППИ РАН, НИУ ВШЭ и Сколтеха
31 января (вторник), 1830, аудитория 615 ИППИ РАН
Дмитрий Яроцкий (ИППИ, Сколтех)
Expressiveness of shallow and deep neural networks: rigorous results
I will discuss some classical and recent rigorous results on the expressive power of neural networks. In particular, I will explain a simple argument showing that deep ReLU networks are more efficient for approximation of smooth functions than shallow ReLU networks.
31.01.2017 | |