How Sigmoid function works in Logistic Regression

Course Q&A Deep Learning Specialization Neural Networks and Deep Learning
,
1

I know the Formulae of Sigmoid function, but how this function actually works with Machine learning.

2

The function basically takes an input whatever might be, there is an infinite range and gives out only 2 outputs either 1 or 0, yes or no!

1 Like
3

The sigmoid function takes a real valued input with range -Inf to +Inf, and compresses the range to real values between 0.0 and 1.0.

Since it uses the exponential function, the sigmoid() function is related the the probability density of a normally distributed variable.

This means it can be viewed as outputting a probability value between 0 and 1.
These map nicely to the logical values of 0 (for false) and 1 (for true), with a decision threshold at 0.5.

So sigmoid tends to be used in classification models.

2 Likes
4

@TMosh,
I would say that the sigmoid function is the Cumulative Distribution Function (CDF) of a logistic distribution with zero mean and the scale one. The logistic distribution resembles the normal distribution in shape but has heavier tails, and it has a simpler CDF than the normal.

1 Like
5

Thanks for the clarification. Statistics is not my best subject.

1 Like