October 2018
Beginner
362 pages
9h 32m
English
Three activation functions that you will commonly see used are the sigmoid, tanh, and Rectified Linear Unit (ReLU) functions. While popular, each of these has a caveat that frequently comes with their use.
During the mid-early 2010s, it was popular to use sigmoids or tanh activation functions in the fully connected layers of a network. Sigmoids have fallen out of fashion, but you might still see them around. They are bounded between the values of 0 and 1 (that is: they can represent any value between that range).
import mathimport numpy as np def sigmoid(x): s_out = [] for item in x: s_out.append(1/(1+math.exp(-item))) return s_out
We can easily plot a sigmoid function in Python to take a look:
Read now
Unlock full access