Computational Intelligence

The Kohonen Network Implementations 245

update_recall_result();

lvq_cur_state = LVQ_NEXT_PATTERN;

}

}

In the LVQ_WEIGHT_STEP_CHANGE state, the winning neuron’s weight change

increments are calculated according to equation 6.11. The state transitions to the

state

LVQ_WEIGHT_CHANGE. The state handling routine is shown here.

static void lvq_weight_step_change (void)

{

int idx_pn;

for (idx_pn = 0; idx_pn < (lvq.arch.layers[0].size) ; idx_pn++)

{ // loop through the connect weights of the current neurons

lvq.arch.layers[1].neurons[lvq.env.winner].delta_w[idx_pn] =

lvq.arch.layers[0].neurons[idx_pn].out -

lvq.arch.layers[1].neurons[lvq.env.winner].w[idx_pn];

lvq.arch.layers[1].neurons[lvq.env.winner].delta_w[idx_pn] ...