Generalized Linear Model

• $p(y; \eta) = b(y) exp(\eta^T T(y) − a(\eta))$
• world

topic

• $y(x) = f(w^Tx+w_0)$. $f$ is called activation function, its inverse is called link function.
• Linear regression:
• Logistic regression: Binomial regression
• Log-linear: Poisson regression