Related Experiment Videos
Nonlinear system modelling via optimal design of neural trees
Yuehui Chen1, Bo Yang, Jiwen Dong
1School of Information science and Engineering, Jinan University, Jiwei Road 106, Jinan, 250022 P. R. China. yhchen@ujn.edu.cn
International Journal of Neural Systems
|April 28, 2004
Summary
This study presents a novel flexible neural tree model optimized using a hybrid learning and evolutionary approach. This method effectively addresses complex problems like time series prediction and system identification.
Area of Science:
- Artificial Intelligence
- Machine Learning
- Computational Neuroscience
Background:
- Neural network models are crucial for complex data analysis.
- Optimizing neural network structures and parameters remains a significant challenge.
- Existing methods may lack flexibility or efficient optimization strategies.
Purpose of the Study:
- To introduce a novel flexible neural tree model.
- To propose a hybrid approach for automatic model optimization.
- To evaluate the model's performance on diverse benchmark problems.
Main Methods:
- A flexible neural tree model is computed using a multi-layer feed-forward neural network.
- A hybrid learning/evolutionary strategy is employed for optimization.
- The modified probabilistic incremental program evolution (MPIPE) algorithm optimizes structure, while a parameter learning algorithm optimizes weights.
Main Results:
- The proposed flexible neural tree model demonstrates effectiveness in function approximation.
- The model shows strong performance in time series prediction tasks.
- System identification problems are successfully addressed, outperforming related methods.
Conclusions:
- The hybrid optimization approach effectively determines optimal neural tree structures and parameters.
- The flexible neural tree model offers a versatile and powerful tool for various machine learning applications.
- This method provides a robust solution for complex modeling tasks.