January 2018
Beginner to intermediate
284 pages
8h 35m
English
The Rectified Linear Unit (ReLU) has become quite popular in recent years. Its mathematical formula is as follows:
Compared to sigmoid and tanh, its computation is much simpler and more efficient. It was proved that it improves convergence by six times (for example, a factor of 6 in Krizhevsky and it's co-authors in their work of ImageNet Classification with Deep Convolutional Neural Networks, 2012), possibly due to the fact that it has a linear and non-saturating form. Also, unlike tanh or sigmoid functions which involve the expensive exponential operation, ReLU can be achieved by simply thresholding activation at zero. Therefore, it ...
Read now
Unlock full access