middle = tanh(Wout * Xtraining);
Ytraining = tanh(Vout * middle);

error_v = 2 * (Ytraining - Dtraining);
grad_v = error_v * middle'; %Calculate the gradient for the output layer

error_w = 2 * ((Vout' * error_v) .* (1 - (middle.^2)));
grad_w = 2 * error_w * Xtraining'; %..and for the hidden layer.

Wout = Wout - learningRate * grad_w; %Take the learning step.
Vout = Vout - learningRate * grad_v; %Take the learning step.