数学科学学院学术报告[2024] 128号

(高水平大学建设系列报告1008号)

报告题目: Computational Principles of Attention Mechanis

报告人：薛云 教授（华南师范大学）

报告时间：2024年11月29日上午10：30-11：30

讲座地点：汇星楼514

报告内容：Unstructured text data contains valuable information, playing a unique role in fields such as public opinion prediction and financial risk management. In recent years, Transformer architectures based on attention mechanisms have gained significant attention in both academia and industry within the field of Natural Language Processing (NLP). These models typically require only pretraining on large amounts of unlabeled text data and can then be fine-tuned on specific downstream tasks with relatively small amounts of labeled data. The Attention mechanism, a specialized structure within machine learning models, is designed to automatically learn and compute the contribution of input data to the output. The lecture is set against the background of NLP, introducing the computational principles of the classical attention mechanism and analyzing its integration with deep learning networks. It aims to engage participants with backgrounds in machine learning and mathematics in discussions on the challenges and methodologies in this field.

报告人简介：薛云，华南师范大学教授，中国人工智能学会青年工作委员会委员，主要从事数据挖掘，自然语言处理等方面的理论和应用研究，在商业智能技术的应用等领域有较深的积累，目前已经发表各类学术期刊和会议论文100余篇，内容涉及数据挖掘、文本大数据分析等方面。申请国家发明专利50余件。

欢迎师生参加！

邀请人：数学科学学院（陈波）

 数学科学学院

                                        2024年11月26日