1 paper
Maxim Milakov, Natalia Gimelshein
The Softmax function is ubiquitous in machine learning, multiple previous works suggested faster alternatives for it. In this paper we propose a way to compute classical Softmax wi…