Improved Matlab-based Gray Wolf Algorithm in Deep Learning Extreme Learning Machine (GWO-DELM) Data Regression Prediction
In this article, we explore how to improve the Matlab-based Gray Wolf algorithm to improve the performance of Deep Learning Extreme Learning Machine (DELM) in data regression prediction problems. We will introduce the basic principle of Gray Wolf Algorithm, and combine DELM framework to realize an improved algorithm. In addition, we also provide the corresponding source code to help readers better understand and practice.
1. Overview of Gray Wolf Algorithm
Gray Wolf Optimization (GWO for short) is an optimization algorithm based on the behavior patterns of gray wolf groups in nature. It simulates the behavioral characteristics of leaders, followers and masterless in gray wolf packs. By simulating these behaviors, the gray wolf algorithm can find the optimal solution and is widely used in function optimization problems.
2. Introduction of DELM Framework
Extreme Learning Machine (Extreme Learning Machine, referred to as ELM) is a fast, simple and effective neural network learning algorithm. It does not need to adjust the weights and biases during the training process, it only needs to randomly generate the connection weights between the input layer and the hidden layer, and then solve the weights of the output layer through the regularization method. DELM is an improved version of ELM, which has better generalization ability and learning speed when dealing with nonlinear problems.
3. GWO-DELM Algorithm Improvement Ideas
In order to improve the performance of GWO-DELM in data regression prediction, we propose the following improvement ideas:
-
Multi-group search strategy:
The original GWO algorithm is extended to multiple groups of gray wolves, and each group searches independently. By introducing multiple groups, the coverage of the search space can be increased and the global search ability can be improved. -
Adaptive adjustment parameters:
In each round of iteration, according to the change of the current fitness value, the parameters of the algorithm are adaptively adjusted. In this way, the algorithm can pay more attention to exploration in the early stage, and pay more attention to local search in the later stage, so as to improve the convergence speed and accuracy of the algorithm. -
Candidate Alternative Selection Strategy