Parameter Tuning in Fuzzy Inference using Neural Network

1

• Parameter Tuning in Fuzzy Inference using Neural
  Network
• CS679 Neural Network
• by Seungsoo Lee
• Computer Science Department
• KAIST

2
Content

• Introduction
• Fuzzy Neural Network Architecture
• Experiment
• Conclusion

3
Introduction

• Approaches to Integrate Fuzzy and Neural Network
• Fuzzy-based approach
• Fuzzy rule search with neural network
• Parameter tuning with neural network
• Fuzzy operator implementation with neural
  network
• NN-based approach
• Fuzzy data input/output
• Parameter Tuning with Neural Network
• Fuzzy rule coarse description of the system
• Neural network automatic learning with error
  correction

4
Fuzzy Neural Network Architecture

• Basic Architecture of Fuzzy Neural Network
• Composed of 5 layers

layer 1
layer 3
layer 4
layer 5
layer 2
defuzzification
input variable
input variable term
inference rule
output variable term
5
Fuzzy Neural Network Architecture
input

• Inference
• input variable
• matching degree with
• input variable term
• inference rule
• matching degree with
• output variable term
• defuzzification

output
6
Fuzzy Neural Network Architecture

• Parameter Tuning
• Based on error backpropagation
• Input variable
• Inference rule
• Output variable

7
Experiment

• Function Approximation
• Target function

8
Experiment

• Function Approximation
• Approximation by fuzzy rule

Input/Output Variable
Function Approximation Result
9
Experiment

• Function Approximation
• Tuning result with neural network

Input/Output Variable
Function Approximation Result
10
Conclusion

• Fuzzy Neural Network
• Fuzzy Rule
• knowledge expression with linguistic terms
• Neural Network
• automatic parameter tuning
• Advantages of each approach can be used