Title :
Training of supervised neural networks via a nonlinear primal-dual interior-point method
Author :
Trafalis, Theodore B. ; Couellan, Nicolas P. ; Bertrand, S?©bastien C.
Author_Institution :
Sch. of Ind. Eng., Oklahoma Univ., Norman, OK, USA
Abstract :
We propose a new training algorithm for feedforward supervised neural networks based on a primal-dual interior-point method for nonlinear programming. Specifically, we consider a one-hidden layer network architecture where the error function is defined by the L2 norm and the activation function of the hidden and output neurons is nonlinear. Computational results are given for odd parity problems with 2, 3, and 5 inputs respectively. Approximation of a nonlinear dynamical system is also discussed
Keywords :
duality (mathematics); learning (artificial intelligence); neural net architecture; nonlinear programming; feedforward supervised neural networks; nonlinear activation function; nonlinear dynamical system; nonlinear primal-dual interior-point method; nonlinear programming; odd parity problems; one-hidden layer network architecture; training algorithm; Computer architecture; Feedforward neural networks; Industrial engineering; Industrial training; Jacobian matrices; Lagrangian functions; Linear programming; Neural networks; Neurons; Nonlinear dynamical systems;
Conference_Titel :
Neural Networks,1997., International Conference on
Print_ISBN :
0-7803-4122-8
DOI :
10.1109/ICNN.1997.614210
