Logistic regression

The logreg model assumes that the classes $y_i$ for every sample $x_i$ are connected via:

whereby $p(x)$ is the probability of a sample with a sample of class $x$ being of class 1, and $1-p(x)$ is the probability of the sample with value $x$ being of class 0

The probability function is thus $p(x)=s(ax+b)$, where $0\le p(s)\le 1$

With two trainable parameters $a$ and $b$, the logistic regression is the combination of the Linear regression and the Sigmoid function

$f(x)=ax+b$ is often referred to as the logit function with value $x$

Logistic functions need to minimise a function called the Log-likelihood cross-entropy function

We evaluate logistic regression models through a Confusion matrix.