Lecture2-1Perceptron Hypothesis Set

Credit Approval Revisited

A simple hypothesis set: Perceptron

For $\mathbf{x} = (x_1, x_2, ..., x_d)$, ‘features of customer’, compute a weight ‘score’ and
- approve credit if $\sum_{i=1}^d w_ix_i>threshold$
- deny credit if $\sum_{i=1}^dw_ix_i<threshold$

$\mathcal{Y}:\{+1 \textrm{good}, -1 \textrm{bad}\}$, 0 ignored -linear formula $h \in \mathcal{H}$ are

$h(\mathbf{x}) = sign\big( (\sum_{i=1}^d w_ix_i) - threshold \big)$ are called perceptron

Vector Form of Perceptron

$$\begin{aligned} h(\mathbf{x}) &= sign\big( (\sum_{i=1}^d w_ix_i) - threshold \big)\\ &=sign\big( (\sum_{i=1}^d w_ix_i) - threshold \times(+1) \big)\\ &=sign\big( \sum_{i=1}^d w_ix_i \big)\\ &=sign(\mathbf{w}^T \mathbf{x}) \end{aligned}$$

Perceptrons in $\mathbb{R}^2$

$$\textrm{Perceptron} \Longleftrightarrow \textrm{Linear Classifiers}$$