One Class Gaussian

  • In one class gaussian, we assume that our data came from Multivariate Normal Distribution
  • The distribution depends on $\mu$ and $\sum$
    • $\mu$ defines where is the center of the distribution
    • $\sum$ defines what will be the shape of the distribution
  • We learn these 2 variables from the sample data, same like the Logistic Regression
  • When we have the learned weight of $\mu$ and $\sum$, then for any new value, $x$ we find the $f(x)$
    • If $f(x)$ is above a threshold, then it is identified as it has come from that class
    • Otherwise anomaly

Pasted image 20231031020238.png
Mutivariate Normal Distribution