initialize equal weights for all samples
Repeat t = 1,…,T
- learn $f_{t}(x)$ with data weights $\alpha_{i}$
- compute weighted error
- compute coefficient
- $\hat{w_{t}}$ is higher when weighted_error is larger
- recomputed weights $\alpha_{i}$
- Normalize weights $\alpha_{i}$
- if $x_{i}$ often mistake, weight $\alpha_{i}$ gets very large
- if $x_{i}$ often correct, weight $\alpha_{i}$ gets very small