Nuril Asyrofiyyah, Endang Sugiharti

Hyperparameter Optimization Using Hyperband in Convolutional Neural Network for Image Classification of Indonesian Snacks

Introduction

Hyperparameter optimization using hyperband in convolutional neural network for image classification of indonesian snacks. Optimize CNN hyperparameters with Hyperband for Indonesian snack image classification, achieving 79.37% test accuracy on 8 classes. Discover optimal dense neurons & learning rate.

Abstract

Abstract. Indonesia is known for its traditional food both domestically and abroad. Several cakes are included in favorite traditional foods. Of the many types of cakes that exist, it is visually easy to recognize by humans, but computer vision requires special techniques in identifying image objects to types of cakes. Therefore, to recognize objects in the form of images of cakes as one of Indonesian specialties, a deep learning algorithm technique, namely the Convolutional Neural Network (CNN) can be used. Purpose: This study aims to find out how the Convolutional Neural Network (CNN) works by optimizing the hyperband hyperparameter in the classification process and knowing the accuracy value when hyperband is applied to the optimal hyperparameter selection process for classifying Indonesian snack images. Methods/Study design/approach: This study optimizes the hyperparameter Convolutional Neural Network (CNN) using Hyperband on the Indonesian cake dataset. The dataset is 1845 images of Indonesian snacks which consists of 1523 training data, 162 validation data and 160 testing data with 8 classes. In training data, the dataset is divided by 82% on training data, 9% validation, and 9% testing. Result/Findings: The best hyperparameter value produced is 480 for the number of dense neurons 2 and 0.0001 for the learning rate. The proposed method succeeded in achieving a training value of 87.53%, for the validation process it was obtained 66.8%, the testing process was obtained 79.37%. Results obtained from model training of 50 epochs. Novelty/Originality/Value: Previous research focused on the application and development of algorithms for the classification of Indonesian snacks. Therefore, optimizing hyperparameters in a Convolutional Neural Network (CNN) using Hyperband can be an alternative in selecting the optimal architecture and hyperparameters.

Review

This paper presents a study on the image classification of Indonesian snacks using Convolutional Neural Networks (CNNs), with a particular emphasis on hyperparameter optimization (HPO) through the Hyperband algorithm. The primary goal is to assess how CNNs perform when their hyperparameters are tuned using Hyperband and to quantify the resulting accuracy for classifying images of Indonesian cakes. The research addresses a practical challenge in computer vision for culturally specific food recognition, demonstrating the application of deep learning to an interesting and relevant dataset. The abstract effectively outlines the core components of the study, including the chosen methodology (CNN with Hyperband HPO), the dataset characteristics (1845 images across 8 classes with an 82/9/9 train/validation/test split), and the key findings. Specific results, such as the identified optimal hyperparameters (480 dense neurons, 0.0001 learning rate) and the achieved accuracies (training: 87.53%, validation: 66.8%, testing: 79.37% over 50 epochs), are clearly stated. However, a significant gap appears in the large disparity between training, validation, and testing accuracy, particularly the low validation score. This suggests potential overfitting or an issue with the validation set's representativeness, which is not addressed in the abstract and warrants further exploration in the full paper. Additionally, while Hyperband is employed, the abstract does not convey the search space or the specific range of hyperparameters explored, which would provide better context for the optimization effort. In conclusion, the study offers a valuable demonstration of applying a modern HPO technique to a specialized image classification problem. The reported test accuracy provides a baseline for recognizing Indonesian snacks using deep learning. While the abstract clearly establishes the contribution of using Hyperband as an alternative HPO strategy, the discrepancy in accuracy metrics points to an area that requires more detailed analysis and discussion within the full manuscript. For future work, it would be beneficial to benchmark the Hyperband-optimized CNN against other HPO methods or advanced CNN architectures to provide a more comprehensive understanding of its performance and efficiency in this domain.

Full Text

Comments

You need to be logged in to post a comment.