Algorithm

Assign each output node $j$ to an existing class label Initialize all weights $w\in[0,1]$ while(network is not converged): Adjust $\eta(t)$ for $x_k$ in dataset: win_j = closest $w_{j}$ to $x_{k}$ for all $j$ for l = 1; i<=n; i++: if x_k.class() == w_j.class: $\Delta w_{j,l}=\eta(t)(x_{k,l}-w_{j,l})$ else: $\Delta w_{j,l}=\eta(t)(x_{k,l}-w_{j,l})$