Automatic Hyperparameter Tuning of Convolutional Neural Networks Using Cheetah Optimizer for Facial Expression Recognition

Automatic Hyperparameter Tuning of Convolutional Neural Networks Using Cheetah Optimizer for Facial Expression Recognition

Mobarake Aftabi1

1) Islamic Azad University, Mashhad Branch Email:

IOI Code : AAHD-ABKKGC
Publication : National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
    Facial expressions are one of the most important communication channels for humans to convey internal emotions. Automatic Facial Expression Recognition (FER) has always attracted researchers attention due to its wide range of applications in human-computer interaction, psychology, healthcare, and security. In this study, a deep convolutional neural network model with a six-layer architecture is designed and implemented. To improve model performance, key hyperparameters including the number of filters (NF), kernel size (KZ), learning rate (LR), and dropout rate are automatically tuned using the Cheetah Optimizer, a metaheuristic algorithm. Experiments on the standard FER2013 dataset show that the proposed model (CNN-CO) increases recognition accuracy from 61% to 73.5%. Furthermore, Precision, Recall, and F1-Score metrics are improved from 0.56, 0.55, and 0.56 to 0.62, 0.62, and 0.61, respectively. The results demonstrate that combining convolutional neural networks with Cheetah metaheuristic optimization is an effective approach for facial expression recognition.

Keywords : facial expression recognition convolutional neural network cheetah optimizer hyperparameter optimization deep learning