Neural Gas with Competitive Hebbian Learning

5.4.5

lambda_i

lambda initial ($\lambda_i$).

lambda_f

lambda final ($\lambda_f$).

epsilon_i

epsilon initial ($\epsilon_i$).

epsilon_f

epsilon final ($\epsilon_f$).

t_max

The simulation ends, if the number of input signals exceed this value (t_max).

edge_i

Initial value for time-dependend edge aging (T_i).

edge_f

Final value for time-dependend edge aging (T_f).

Edges are removed with an age larger than the maximal age T(t) whereby

$$T(t) = T_i(T_f/T_i) ^ {t/t_{\rm max}}.$$

The reference vectors are adjusted according to

$$\Delta \mbox{\bf w}_i = \epsilon(t) \cdot h_\lambda(k_i(\mbo... ...$\xi$},{\cal A})) \cdot (\mbox{\boldmath$\xi$}- \mbox{\bf w}_i)$$

with the following time-dependencies:

$$\lambda(t) = \lambda_i (\lambda_f/\lambda_i)^{t/t_{\rm max}}$$

$$\qquad\epsilon(t) = \epsilon_i(\epsilon_f/\epsilon_i)^{t/t_{\rm max}}$$

$$h_\lambda(k) = \exp(-k/\lambda(t)).$$