Algorithm 2.

(Training and prediction with the logistic regression classifier)

Training (model generation) stage
1.	Initialize wm = 0
2.	Estimate log-odds ratio wm using maximum likelihood estimation
	wm = argmax l(w), where log-likelihood
	$l\left(w\right)=\sum _{i=1}^n\left(y_i\log p(x_i,w)+\frac{(1-y_i)}{2}\log (1-p(x,w)\right)$ and
	$p(x,w)=\mathit{Pr}(Y=-1,X=x)=\frac{1}{1+e^{-\left(x^{T}w\right)}}$
Prediction (testing) stage
1.	Calculate posterior probability Pr(Y = −1|X = x) using $p(x,w)\frac{1}{1+e^{-\left(x^T{w}_m\right)}}$
2.	$\text{Predict}y=\left\{\begin{array}{cc}\hfill -1,\hfill & \text{if}\mathit{Pr}(Y=-1\mid X=x)>\mathit{Pr}(Y=1\mid X=x)\hfill \\ \hfill +1,\hfill & \text{otherwise}\hfill \end{array}\right\}$