Adaptive higher-order feedforward neural networks
Shuxiang Xu, Ming Zhang · 2003
In this paper we study the approximation capabilities of an adaptive higher-order feedforward neural network (AHFNN) with a neuron-adaptive activation function. A learning algorithm is derived to tune the free parameters in the neuron-adaptive activation function as well as connection weights between neurons. Simulation results show that the proposed AHFNN presents several advantages over traditional neuron-fixed networks such as increased flexibility, much reduced network size, faster learning, and lessened approximation errors. Experiments also reveal that AHFNN is especially superior in financial data simulation and financial prediction.