Mr. ATMAJA, Bagus Tris (3rd-year doctoral student in Akagi lab of Human Life Design Area) received Best Student Paper and Presentation Award in the 1st Biennial International Conference on Acoustic and Vibration 2020 (ANV2020).

The Conference is organized by The Vibration and Acoustics Laboratory (VibrasticLab), Engineering Physics Department, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia. ANV 2020 is technical conference on acoustics, vibration, and its applications. The theme of the Conference is "Sound of Indonesia" and it was held Online. We expect a broad program of sessions on variety of acoustics and vibration topics, and its application. This provides a great opportunity and arena for industry participation and exhibition.

*Reference：ANV2020

■Date Awarded
November 24, 2020

■Title
Evaluation of Error and Correlation-Based Loss Functions For Multitask Learning Dimensional Speech Emotion Recognition

■Authors
Bagus Tris Atmaja and Masato Akagi

■Abstract
The choice of a loss function is a critical part in machine learning. This paper evaluates two different loss functions commonly used in regression-task dimensional speech emotion recognition ― error-based and correlation-based loss functions. We found that using correlation-based loss function with concordance correlation coefficient (CCC) loss resulted in better performance than error-based loss functions with mean squared error (MSE) and mean absolute error (MAE). The evaluations were measured in averaged CCC among three emotional attributes. The results are consistent with two input feature sets and two datasets. The scatter plots of test prediction by those two loss functions also confirmed the results measured by CCC scores.

■Comment
I would like to thank Akagi-sensei for supervising me and MEXT for funding my research. ANV 2020 is the first biennial conference on acoustic and vibration. This conference is planned to be held in Bali, Indonesia. However, due to Covid-19, this conference was held virtually via Zoom. In this conference, I presented my word on evaluation of correlation-based loss function for dimensional speech emotion recognition. Since the goal is performance measured in correlation-based score, it is reasonable to also choose a correlation-based loss function. The experimental results show the usefulness of using correlation-based loss function over error-based loss function.

December 24, 2020