Dropout Forward Layer

The forward dropout layer computes function y = B(r) * x / r for input argument x, where B(r) is a Bernoulli random variable with parameter r.

Problem Statement

Given a p-dimensional tensor X of size n1 x n2 x ... x np, the problem is to compute p-dimensional tensors M = (mi1...ip) and Y = (yi1...ip) of size n1 x n2 x ... x np, where:

mi1...ip = B(r) / r

yi1...ip = mi1...ip * xi1...ip

For more complete information about compiler optimizations, see our Optimization Notice.
Select sticky button color: 
Orange (only for download buttons)