Logistic Regression

Classification of $y \in \{0,1\}$

$$p(y|x,\theta) = \text{Ber}(y | \sigma(w^{T}x + b))$$

• logit $a$: $w^{T}x + b$
• sigmoid:

$$\sigma(a) = \frac{1}{1 + e^{-a}} = p(y = 1 | x,\theta)$$