Focal Modulation Networks for Interpretable Sound Classification
It addresses interpretability for audio domain applications, which is an incremental advancement as it adapts an existing method to a new domain.
The paper tackles the problem of interpretability-by-design in audio classification by applying focal modulation networks to environmental sound classification on the ESC-50 dataset, achieving higher accuracy and interpretability than a similarly sized vision transformer and competitive performance against a post-hoc interpretation method.
The increasing success of deep neural networks has raised concerns about their inherent black-box nature, posing challenges related to interpretability and trust. While there has been extensive exploration of interpretation techniques in vision and language, interpretability in the audio domain has received limited attention, primarily focusing on post-hoc explanations. This paper addresses the problem of interpretability by-design in the audio domain by utilizing the recently proposed attention-free focal modulation networks (FocalNets). We apply FocalNets to the task of environmental sound classification for the first time and evaluate their interpretability properties on the popular ESC-50 dataset. Our method outperforms a similarly sized vision transformer both in terms of accuracy and interpretability. Furthermore, it is competitive against PIQ, a method specifically designed for post-hoc interpretation in the audio domain.