Neural network structures with constant weights to implement dis-jointly removed non-convex (DJRNC) decision regions: part a - properties, model, and simple case

Authors:
Che-Chern Lin
Affiliations:
National Kaohsiung Normal University, Department of Industrial Technology Education, Kaohsiung, Taiwan, R.O.C.
Venue:
NN'07 Proceedings of the 8th Conference on 8th WSEAS International Conference on Neural Networks - Volume 8
Year:
2007

Citing 5
Cited 1

Applying SP-MLP to complex classification problems

Pattern Recognition Letters
The constraint based decomposition (CBD) training architecture

Neural Networks
Classification ability of single hidden layer feedforward neural networks

IEEE Transactions on Neural Networks
A constructive algorithm to solve “convex recursive deletion” (CoRD) classification problems via two-layer perceptron networks

IEEE Transactions on Neural Networks
A two-layer paradigm capable of forming arbitrary decision regions in input space

IEEE Transactions on Neural Networks

Empirical determination of sample sizes for multi-layer perceptrons by simple RBF networks

WSEAS Transactions on Computers

Quantified Score

Hi-index	0.00

Visualization

Abstract

This is a series of studies to discuss the partitioning capabilities of multi-layer perceptrons on dis-jointly removed non-convex (DJRNC) decision regions. There are two papers proposed in the series of studies including part A and part B. In part A, we propose a network structure to implement DJRNC decision regions using multi-layer perceptrons. In the proposed structure, all weights and the parameters of the activation functions are pre-determined when a DJRNC decision region is established. No constructive algorithm is needed for implementing the DJRNC decision regions and each weight determined by this paper is either 1 or -1. This makes the hardware implementations of the proposed network structures easy. Three cases are discussed in this paper including single, nested, and disconnected decision regions. The first case is shown in part A, and the rest of two are demonstrated in part B. We also provide three multi-layer perceptrons to implement the three decision regions and prove the implementation feasibilities of the proposed model.