Definition

Combines both Hamming Network and MaxNet, in sequence, then brings the winner node closer to the input node. # Algorithm 1. Initialize random weights and learning rate $\eta$. 2. Find the winning cluster $j$ (of $m$ clusters) which is the cluster closest to the input vector $x=[x_{1}\ldots x_{n}]$ (minimize the squared distance of each cluster’s weight with respective element of vector) with each cluster’s associated weight $w_{j,i}$ $$\min_{j=1,..m} D(j)=\min_{j=1,..m}\left\{\sum\limits_{i=1}^{n}(x_{i}-w_{j,i})^{2}\right\}$$ 3. Update each $w_{j,i}$ for the chosen $j$ like so: $$\Delta w_{j,i}=\eta(t)(x_{i}-w_{j,i}), \ w_{j,i}=w_{j,i}-\Delta w_{j,i}$$for $i\in\{1,\ldots,n\}$ 4. Repeat until network converges.