Title :
Neural Network Learning Without Backpropagation
Author :
Wilamowski, B.M. ; Hao Yu
Author_Institution :
Dept. of Electr. & Comput. Eng., Auburn Univ., Auburn, AL, USA
Abstract :
The method introduced in this paper allows for training arbitrarily connected neural networks, therefore, more powerful neural network architectures with connections across layers can be efficiently trained. The proposed method also simplifies neural network training, by using the forward-only computation instead of the traditionally used forward and backward computation.
Keywords :
Hessian matrices; Jacobian matrices; forward chaining; learning (artificial intelligence); neural net architecture; Hessian matrix; Jacobian matrix; Levenberg Marquardt algorithm; forward-only computation; gradient vector; neural network architecture; neural network learning; neural network training; Artificial neural networks; Backpropagation; Frequency modulation; Jacobian matrices; Neurons; Forward-only computation; Levenberg–Marquardt algorithm; neural network training; Algorithms; Artificial Intelligence; Computer Simulation; Mathematical Computing; Neural Networks (Computer); Neurons; Software Design; Time Factors;
Journal_Title :
Neural Networks, IEEE Transactions on
DOI :
10.1109/TNN.2010.2073482
