Hands-On ROS for Robotics Programming

Logistic regression

Logistic regression is used in classification problems, a very common type in ML. In this case, we try to predict a binary classification such as pass/fail, win/lose, alive/dead, or healthy/sick. This algorithm can be understood as a special case of regression, where the predicted variable is categorical, that is, it can only take a finite set of values (two if it is a binary classification). The underlying model is a probability function and, given a value of the independent variable, if the resulting probability is greater than 50%, we predict pass, win, alive, or healthy, and if lower, the prediction is the other category, that is, fail, lose, dead, or sick.