Metaheuristics Approach for Hyperparameter Tuning of Convolutional Neural Network

Metaheuristics Approach for Hyperparameter Tuning of Convolutional Neural Network

convolutional neural network;hyperparameter;metaheuristics;ACO;GA. HS

Deep learning is an artificial intelligence technique that has been used for various tasks. The performance of deep learning is determined by its hyperparameter, architecture as well as training (connection weight and bias). Finding the right combination of those aspects isvery challenging. Convolution Neural Networks (CNN) is a deep learning method that is commonly used for image classification. It has many hyperparameterstherefore tuning its hyperparameter is difficult. In this research, a metaheuristics approach is proposed to optimisethe hyperparameter of convolution neural networks. Threemetaheuristics methodsare used in this research, ant colony optimization (ACO,) genetic algorithm (GA)and Harmony Search (HS). Themetaheuristics methods are usedtofind the best combination of8 hyperparameterswith 8 optionseach which creates1.6. 107 ofsolution space.The solution space is too big to explore using manual tuning. The Metaheuristics method willbring benefitsin termsof finding solutions in the search space more effectively and efficiently.The performance of the metaheuristics methodsisevaluated using MNIST datasets. The experiment resultsshow that theaccuracy of ACO, GA and HS are 99,7%, 97.7% and 89,9% respectively. The computationaltime for the ACO, GA and HS algorithmsare 27.9 s, 22.3 s and 56.4 srespectively. It showsthat ACO performsthe best among the three algorithmsin termsof accuracy however its computational time is slightly longer than GA.The experiment results revealthat themetaheuristic approach is promising for the hyperparameter tuning of CNN.Future research can be directed to solve larger problems or enhancethemetaheuristics operator to improve its performance

Hindriyanto Dwi Purnomo1, Tad Gonsalves2, Evangs Mailoa3, Fian Yulio Santoso4, Muhammad Rizky Pribadi

https://jurnal.iaii.or.id/index.php/RESTI/article/view/5730/935

Departmentof Information Technology, Facultyof Information Technology, Satya Wacana Christian University, Salatiga, Indonesia

01-06-2024

FAJAR BAGUS W

PDF

ENGLISH

TEXT