Improved Gray Wolf Algorithm Based on African Vulture Multi-Strategy and Its Application in Channel Estimation

The Grey Wolf Optimizer (GWO) is known for its simple structure, easy implementation, and few control parameters, but it has problems such as slow convergence, local optimization of capture, and late reduction of population diversity. To address these shortcomings, this paper proposes fusing three different strategies. Firstly, the African vulture algorithm is combined with the GWO to exploit the fast convergence speed and high accuracy of the African vulture algorithm in finding the best solution. Secondly, the leading wolf guidance strategy is introduced to increase population diversity. Finally, adaptive weights are added to dynamically adjust the search step size and balance the global and local search ability of the GWO. The proposed algorithm is validated through testing on 23 benchmark test functions, demonstrating its superior performance. For millimeter-wave large-scale MIMO systems, a deep convolutional neural network is used for channel estimation, using the correlation between space and frequency to simultaneously input the corrupted channel matrix of adjacent subcarriers into the convolutional neural network. Furthermore, we address the drawbacks of manually setting the hyperparameters of convolutional neural networks, which may lead to overfitting and large errors. Therefore, we propose to use the improved grey Wolf algorithm (AVGWO) of African vulture to find the hyperparameters of the convolutional neural network model and improve the accuracy of the model prediction channel.