Mathematics-Inspired Models: A Green and Interpretable Learning Paradigm for Multimedia Computing DOI Open Access
Lei Gao, Kai Liu, Zheng Guo

et al.

ACM Transactions on Multimedia Computing Communications and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: March 3, 2025

The advances of machine learning (ML), and artificial intelligence (AI) in general, have attracted unprecedented attention intelligent multimedia computing many other fields. However, due to the concern for sustainability black-box nature ML models, especially deep neural networks (DNNs), green interpretable learnings been extensively studied recent years, despite suspicions on effectiveness, subjectivity interpretability complexity. To address these concerns suspicions, this paper starts with a survey discoveries learning, then presents mathematics-inspired (M-I) models. We will demonstrate that M-I models are numerous properties. Finally we present several examples multi-view information both static image-based dynamic video-based tasks methodology promises plausible sustainable path natural evolution ML, which is worth further investment in.

Language: Английский

Mathematics-Inspired Models: A Green and Interpretable Learning Paradigm for Multimedia Computing DOI Open Access
Lei Gao, Kai Liu, Zheng Guo

et al.

ACM Transactions on Multimedia Computing Communications and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: March 3, 2025

The advances of machine learning (ML), and artificial intelligence (AI) in general, have attracted unprecedented attention intelligent multimedia computing many other fields. However, due to the concern for sustainability black-box nature ML models, especially deep neural networks (DNNs), green interpretable learnings been extensively studied recent years, despite suspicions on effectiveness, subjectivity interpretability complexity. To address these concerns suspicions, this paper starts with a survey discoveries learning, then presents mathematics-inspired (M-I) models. We will demonstrate that M-I models are numerous properties. Finally we present several examples multi-view information both static image-based dynamic video-based tasks methodology promises plausible sustainable path natural evolution ML, which is worth further investment in.

Language: Английский

Citations

0