报 告 人 |
Xue Li, Professor, School of Electrical Engineering and Computer Science, The University of Queensland, Australia |
报 告 内 容 简 介 |
报告内容简介： Data-driven AI research covers the entire life cycle of data, from metadata modelling, data collection, quality management, information dissemination, false information, misinformation, disinformation, multimodal data fusion, privacy protection, machine learning and training, data mining and pattern recognition, knowledge graph generation and management, and the effective application of data in AI-enabled medical applications. This talk explores the collection, analysis, processing, and other issues of multimodal data engineering, focusing on the contributions in medical applications made by our team in this field. With the development of generative AI, machine learning, including large language models (LLM), the application of artificial intelligence faces many challenges and pitfalls. However, the processing of multimodal data also faces many challenges in algorithm design and optimization and semantic understanding. In terms of algorithm design and optimization, multimodal data needs to solve the alignment of semantic units of different modalities, Data linkage for different syntactic units, many-to-many, and the complexity and consistency at different target granularities in feature engineering. In order to deal with these research problems, our research team proposed some optimization strategies and algorithm frameworks, such as the relationship between feature alignment of multi-granularity, multi-level, and multi-time intervals and knowledge graph improvement, and contrastive learning on tabular data. This report takes data expression as the starting point when introducing the problem and introduces some synthetic data and simulation data methods in current data engineering. |
浙公网安备 33100202001351号
