Algorithm 2 Training the best classification line

Input: predicted classification line and observed measure point set ${\{x_i,y_i\}}_{i=1}^l$ at time $k+1$; Output: best classification line at moment $k+1$;  1: Input the predicted classification line and observed measure point set  ${\{x_i,y_i\}}_{i=1}^l$ at time $k+1$;  2: Classify the observed measure point set ${\{x_i,y_i\}}_{i=1}^l$ by using the predicted  classification line according to Equation (19), which will get the classified  sampling point set ${\{x_i,y_i,z_i\}}_{i=1}^l$ with the binary class labels $z_i\in \{-1,+1\}$;  3: Substituting ${\{x_i,y_i,z_i\}}_{i=1}^l$ into Equation (23) to compute the parameter  $\mathsf{\omega }$ and b, and finally get the best classification line at moment $k+1$.