S. Gardner, R. Lamb, and J. Paxton (USA)
Neural Network, Backpropagation, Feedback Control,Fractional Calculus
The widely used backpropagation algorithm based on stochastic gradient descent suffers from typically slow convergence to either local or global minimum error. This backpropagation algorithm bears great resemblance to a classic proportional integral derivative (PID) control system. Fractional calculus shows promise for improving stability and response in feedback control through the use of non-integer order derivatives and integrals. In this paper, the application of fractional calculus to backpropagation is explored as a means for improving speed and performance.
Important Links:
Go Back
