计算机学科前沿论坛之二
报告题目：Deep Spectral Kernel Networks
报告人：薛晖
报告时间：2023年11月11日下午2：00
地点：计算机楼A501
报告摘要：Since the release of ChatGPT, generative AI has caught the attention of many individuals, corporations and governments. It is a very disruptive technology that is already changing how people learn and work with producing high quality content, specifically text, image and audio, which often exhibit non-stationary characteristics. Non-stationary and non-monotonic spectral kernels effectively break the local limitation and provide a new idea for in-depth analysis, understanding and prediction of dynamic non-stationary data. In this talk, we will systematically present our latest research works about the non-stationary spectral kernel: 1) the framework of deep spectral kernel networks (DSKN); 2) how to solve the optimization dilemma of DSKN; 3) how to improve the representation capability of DSKN; 4) a solid application method for time series domain adaptation based on DSKN. Finally, we will discuss profound thinking on further development of DSKN in generative AI.
报告人简介：薛晖，东南大学计算机科学与工程学院、软件学院、人工智能学院教授，博士生导师。主要研究领域包括机器学习与模式识别。在TPAMI、NeurIPS、AAAI、IJCAI、CVPR、ACM MM等重要国际期刊和领域顶级会议上发表论文70余篇，主持多项国家自然科学基金项目和江苏省自然科学基金项目。现任中国计算机学会人工智能与模式识别专业委员会委员、中国人工智能学会机器学习专业委员会委员、江苏省人工智能学会机器学习专业委员会副主任、江苏省人工智能学会智能系统与应用专业委员会副主任等。
