Algorithm 2 Training algorithm for an Optimized L-Tree.
	Prepare ${\mathrm{W}}_{best}$: the best SOM weights to classification the object $I{G}_{best}$: the information gain when the SOM weights has best classification ability
	Build an optimized L-Tree ($\overline{\overline{\mathrm{X}}}$) 1:$\mathbf{for}$ $t=1:T$ 2:  Extract random local-area position $(x_{st},y_{st})$ 3:  normalize by subtracting the mean, ${\overline{\overline{\mathrm{x}}}}^{{}^{\prime }}$ 4:  Learning SOM weights ${\mathrm{W}}_t$ through the Node learning ($\overline{\overline{\mathrm{x}}}$) 5:  Split $\overline{\overline{\mathrm{X}}}$ into subset ${\overline{\overline{\mathrm{X}}}}_1,{\overline{\overline{\mathrm{X}}}}_2,\dots ,{\overline{\overline{\mathrm{X}}}}_K$ according to the similarity with K trained nuerons. 6:  Calculate parent and child entropies $H(\overline{\overline{\mathrm{X}}})$ when the dataset is classified with $\mathrm{W}$. 7:  Calculate Information gain ($I{G}_t(\overline{\overline{\mathrm{X}}},W_t)$) from the entropy. 8:  if $I{G}_t>I{G}_{best}$ 9:    ${\mathrm{W}}_{best}={\mathrm{W}}_t$ 10:    $I{G}_{best}=I{G}_t$ 11:  End if 12:$\mathbf{End}\mathbf{for}$ 13:for $j=1,2,\dots ,K$ check a $\mathbf{terminal}\mathbf{criteria}$ 14:If $\mathbf{6}$ satisfying stop the node learning, else then Build an optimized L-Tree (${\overline{\overline{\mathrm{X}}}}_j$).