Novel back propagation optimization by Cuckoo Search algorithm.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Neural Networks

Background:

• Traditional Back Propagation (BP) networks suffer from slow training, local minima, and sensitivity to initial weights and bias.
• These limitations hinder the efficiency and accuracy of neural network models in complex tasks.

Purpose of the Study:

• To propose an improved Back Propagation (BP) network, termed CSBP, optimized using Cuckoo Search (CS).
• To enhance the performance of BP networks by addressing their inherent limitations through metaheuristic optimization.
• To evaluate the predictive capabilities of the CSBP model using real-world data.

Main Methods:

• The Cuckoo Search (CS) algorithm is employed to simultaneously optimize the initial weights and bias of the BP network.
• The proposed CSBP model is trained and tested using wine data for predictive modeling.
• Performance comparison is conducted against the standard BP algorithm and the General Regression Neural Network (GRNN).

Main Results:

• The CSBP model demonstrates superior performance compared to the basic BP network.
• CSBP shows competitive results when compared with the General Regression Neural Network (GRNN).
• Parameter sensitivity analysis is performed to optimize the implementation of CSBP.

Conclusions:

• The proposed CSBP model effectively mitigates the disadvantages of traditional BP networks.
• CSBP offers a promising approach for improving the accuracy and efficiency of neural network-based predictions.
• Further parameter tuning can enhance the practical application of CSBP in various domains.