Algorithm 1 Self-Organizing Tree Map for Incremental Feature Extraction

Input: Feature vectors: $\left\{f_1,\dots ,f_N\right\}$; maximum number of iterations: $MaxItr$; maximum number of clusters: $MaxCls$ Initialize the root node $C_1\left(1\right)$ with a randomly selected feature vector repeat {Initialize $r$ = 1 and $k$ = 1} Take another feature vector $f\left(r\right)$ randomly and find the closet node $C_k^{*}\left(r\right)$ based on the minimum Euclidean distance $d_k^{*}$. Calculate $H\left(r\right)$ using Equation (4). Calculate $\alpha \left(r\right)$ using Equation (5). If$d_k^{*}\le H\left(r\right)$then update node $C_k^{*}\left(r\right)$ using Equation (6) $$C_k^{*}\left(r+1\right)=C_k^{*}\left(r\right)+\alpha \left(r\right)\left[f\left(r\right)-C_k^{*}\left(r\right)\right]$$ (6) else increment $k$ and initialize the new node $C_1\left(k\right)$ with $f\left(r\right)$ Increment $r$ until$r$ is equal to $MaxItr$ or $k$ is equal to $MaxCls$. Output: Cluster centroids $C_1,\dots ,C_K$ and their associated feature vectors ${\left\{v_n\right\}}_{n=1:m_k}$.