Logit

I see this a lot in Deep Learning.

In statistics, the logit function is the quantile function associated with the standard logistic distribution. It has many uses in data analysis and machine learning, especially in data transformations.

$\mathrm{logit}(p)=\log (\frac{p}{1-p})$ where

• $p$ = probability
• $\frac{p}{1-p}$ = corresponding odds

Logistic Regression

In deep learning, I think the logit are the raw unnormalized values before you apply the Softmax.