Step 3. Calculate the error of each node Ej using eqn : dj - yj if j is an output node ej —
where dj is a desired output of node j. • Step 4. Relax the error-propagation of the RBP to find the fixed point zj using eqn :
td 'a ) = - Zj(t) + X a\ X wuyi(t) + xt (t) ) WjkZk(t)
where a9 (.) is the time derivative of the activation function.
Step 5. Update the weights using eqns  and : Aw,j = r)8jyi 
Various models were presented to learn time-series data using recurrent algorithms. Connectivity among the computation nodes was accordingly designed to extract information on ecological time processes and to predict future development of ecological systems.
Was this article helpful?