A novel Taguchi-based approach for optimizing neural network architectures: application to elastic short fiber composites

TL;DR

This paper introduces a Taguchi-based method to efficiently optimize neural network hyperparameters for predicting elastic properties of short fiber composites, reducing computational effort while improving accuracy.

Contribution

It applies the Taguchi design of experiments to neural network hyperparameter optimization, demonstrating enhanced efficiency and performance in material property prediction.

Findings

Taguchi method reduces hyperparameter tuning time.

Optimized neural network achieves higher prediction accuracy.

Significant computational resource savings observed.

Abstract

This study presents an innovative application of the Taguchi design of experiment method to optimize the structure of an Artificial Neural Network (ANN) model for the prediction of elastic properties of short fiber reinforced composites. The main goal is to minimize the required computational effort for hyperparameter optimization while enhancing the prediction accuracy. Utilizing a robust design of experiment framework, the structure of an ANN model is optimized. This essentially is the identification of a combination of hyperparameters that yields an optimal predictive accuracy with the fewest algorithmic runs, thereby achieving a significant reduction of the required computational effort. Our findings demonstrate that the Taguchi method not only streamlines the hyperparameter tuning process but also could substantially improve the algorithm's performance. These results underscore the potential of the Taguchi method as a powerful tool for optimizing machine learning algorithms, particularly in scenarios where computational resources are limited. The implications of this study are far-reaching, offering insights for future research in the optimization of different algorithms for improved accuracies and computational efficiencies.