Algorithm 1 Outline of Real AdaBoost.

Input: Training data samples, (si, li), i = 1,2,…, N; si ∈ ℝn, li ∈ {+1, −1}

Initialization: Set uniform sample weights, $w_1(i)=\frac{1}{N}$.

Iteration: For t = 1, 2, …, T, repeat the following steps: Train the weak classifier to obtain a class probability estimate, P(l = 1∣s) ∈ [0, 1], using the weighted samples. Set $h_t(s)=\frac{1}{2}log\frac{P(l=1|s)}{P(l=-1|s)}$. Update the sample weights: $w_i^{(t+1)}=w_i^{(t)}exp(-y_i{h}_t(s_i))$, i = 1, 2, …, N. Normalize the weight distribution: $w_i^{(t+1)}=\frac{w_i^{(t+1)}}{\sum _{i=1}^N{w}_i^{(t+1)}}$.

Output: The final strong classifier, $H(s)=\mathit{\text{sign}}\left(\sum _{t=1}^T{h}_t(s)\right)$