The state of the art of non-linearity is to use ReLU instead of sigmoid function in deep neural network, what are the advantages?

I know that training a network when ReLU is used would be faster, and it is more biological inspired, what are the other advantages? (That is, any disadvantages of using sigmoid)?

Best answer in stackexchange:

Two additional major benefits of ReLUs are sparsity and a reduced likelihood of vanishing gradient. But first recall the definition of a ReLU is h=max(0,a)h=max(0,a) where a=Wx+ba=Wx+b.

One major benefit is the reduced likelihood of the gradient to vanish. This arises when a>0a>0. In this regime the gradient has a constant value. In contrast, the gradient of sigmoids becomes increasingly small as the absolute value of x increases. The constant gradient of ReLUs results in faster learning.

The other benefit of ReLUs is sparsity. Sparsity arises when a≤0a≤0. The more such units that exist in a layer the more sparse the resulting representation. Sigmoids on the other hand are always likely to generate some non-zero value resulting in dense representations. Sparse representations seem to be more beneficial than dense representations.

Reference: http://stats.stackexchange.com/questions/126238/what-are-the-advantages-of-relu-over-sigmoid-function-in-deep-neural-network

ReLU

ReLU的全称是rectified linear unit。上面的回答基本上涵盖了它胜过sigmoid function的几个方面：

1. faster
2. more biological inspired
3. sparsity
4. less chance of vanishing gradient (梯度消失问题)

早期使用sigmoid或tanh激活函数的DL在做unsupervised learning时因为 gradient vanishing problem 的问题会无法收敛。ReLU则这没有这个问题。