A Survey on Deep Learning for Multimodal Data Fusion.

Area of Science:

• Computer Science
• Data Science
• Artificial Intelligence

Background:

• Heterogeneous networks generate vast multimodal big data, characterized by high volume, variety, velocity, and veracity.
• Traditional data fusion methods struggle with the complexity and intermodality of this data.
• Multimodal big data offers rich intermodality and cross-modality information crucial for advanced analytics.

Purpose of the Study:

• To survey pioneering deep learning models for multimodal big data fusion.
• To provide a foundational understanding of deep learning fusion methods for diverse audiences.
• To stimulate the development of novel deep learning techniques for multimodal data fusion.

Main Methods:

• Review and summarization of representative deep learning architectures for multimodal fusion.
• Analysis of current pioneering deep learning models applied to multimodal data fusion.
• Identification of challenges and future research directions in the field.

Main Results:

• Key deep learning architectures fundamental to multimodal fusion are identified.
• Pioneering deep learning models for fusing multimodal big data are cataloged.
• Current limitations and promising future research avenues are outlined.

Conclusions:

• Deep learning offers powerful solutions for fusing complex multimodal big data.
• Further research is needed to address existing challenges and unlock new fusion techniques.
• This review serves as a guide for understanding and advancing multimodal deep learning fusion.