学术报告:Recent Applications of Probability Models in Deep Learning 概率模型在深度学习中的应用

题目:Recent Applications of Probability Models in Deep Learning 概率模型在深度学习中的应用

主讲人:Richard Xu,徐亦达,悉尼科技大学, 副教授, 机器学习与数据分析实验室主任

日期:2018年10月30日(星期二)

时间:上午10:00 - 11:30

地点:数据科学与计算机学院 A101

主持:吴贺俊 副教授

摘要:

In this talk, we review three recent papers where traditional probability methodologies and Deep Learning methods have been helping each other. Firstly, we discuss how classic Expectation Maximization (EM) has been used to assist recently popular matrix capsule’s routing algorithm, Secondly, we discuss how LSTM model have been helping to tune the parameters of a Kalman Filter; Lastly, we discuss how Determinantal Point Process (DPP) have been used to perform better Deep Neural network compressions.

许多深度学习技术与一些贝叶斯方法之间存在千丝万缕的联系。在本次报告中,我们将回顾最近的三篇关于如何将传统的概率方法和深度学习方法相互帮助的文章。首先,我们将讨论经典期望最大化(EM)如何用于辅助最近流行的矩阵胶囊网路中的动态路由算法。其次,我们来讨论LSTM模型是如何帮助调整卡尔曼滤波器的参数; 最后我们将讨论如何使用行列式点过程来执行更好的深度神经网络压缩工作。

个人介绍:

Richard Yi Da Xu is an associate professor in Machine Learning at University of Technology Sydney and a Director of Industry Analytics and Visualization. He leads team of 30 people, include postdoc, PhD students and data engineers; His main research is in probabilistic machine learning, deep leaning and computer vision. He published at many International conference and journals including AAAI, IJCAI, and AISTATS, IEEE-(TNNLS, TIP, TKDE, and T-Cybernetics). Since 2009, he published 1000+ slides of PhD training material in machine learning as well many online ML videos. His team has collaborated with many Australian and Chinese industries, including banks, e-commerce, government, utilities and law firms; He established a Deep Learning Sydney meetup group, which currently has 3200+ members. He is a sole representative for Australia to attend ISO JTC1 SC42 Artificial Intelligence’s first plenary.  

徐亦达,悉尼科技大学副教授,机器学习与数据分析实验室主任。主要从事机器学习方面的 研究工作,包括深度学习,数据分析和计算机视觉等。他的文章发表在 AAAI, IJCAI, AISTATS, IEEE-TNNLS, TIP, TKDE, Cybernetics 等国际会议和期刊上。他从 2009 年来一直致力于撰写机器 学习公开课讲义与录制中文机器学习视频。他带领的团队与许多澳大利亚和国际著名的大型 机构,包括银行,政府部门,电网,律师事务所拥有长期人工智能研究合作。他创立并组织了悉 尼深度学习见面会, 目前拥有会员人数 3200 人。他作为澳大利亚的唯一代表参加 2018 年 4 月 ISO JTC 1/SC42 国际人工智能标准第一次全体会议。