Algorithm 1 Estimation of RBFN centers via IF

Input   •$X_{train}$         % Data to train the RBFN;   •$\alpha$         % Learning rate constant;   •$n\_center$         % Number of initial candidates;   •$\delta$         % Threshold of Information Potential;   •$\beta$         % Minimum distance between candidates; $max\_epochs$         % Maximum epochs;   •$\gamma$         % Constant of convergence;      Output   •$C_{cluster}$         % Cluster centers. 1: procedure IF_algorithm($X_{train}$,$\alpha$,$n\_center$,$\delta$,$\beta$,$max\_epochs$,$\gamma$) 2: 3:     % $P_{if}(·)$ calculates the information potential. 4:     % $F_{if}(·)$ calculates the information force. 5:     for $i\in 1:n\_center$ do          % Raffle the candidates 6:        $C_{cand}\left(i\right)=random\left(X_{train}\right)$; 7:         if $P_{if}\left(C_{cand}\left(i\right)\right)<\delta$ then 8:            Eliminate $C_{cand}\left(i\right)$; 9:         end if 10:     end for 11: 12:     $s=0;$ 13:     while ($s<max\_epochs$) and (not all $C_{candidate}$ are eliminated or converged) do 14:         % Eliminate points to close each other 15:         for i and $j\in 1:n\_center$ do 16:            if $\left|\right|(C_{cand}\left(i\right)-C_{cand}\left(j\right))\left|\right|<\beta$ then 17:                Eliminate $C_{cand}\left(i\right)$ 18:            end if 19:         end for 20: 21:         for $i\in 1:n\_center$ do             ▹ Update the center candidate. 22:            $C_{cand}(i,s+1)=C_{cand}(i,s)+\alpha \ast e^{\left|\right|F_{if}(i,s)\left|\right|}\ast F_{if}(i,s)$ 23:            if $|e^{\left|\right|F_{if}(i,s+1)\left|\right|}-e^{\left|\right|F_{if}(i,s)\left|\right|}|<\gamma$ then 24:                $C_{cand}\left(i\right)$ converge! 25:            end if 26:         end for 27:         $s=s+1$ 28:     end while 29:     $C_{cluster}=\cup C_{cand}$ that converge. 30: return $C_{cluster}$ 31: 32: end procedure