Machine learning, once considered a branch of artificial intelligence (AI), has recently emerged as an independent discipline itself. Machine learning studies the problem of how would a computer learn from data, to generate useful rules and improve its behavior. This problem is central to AI, since to be considered intelligent, the computer cannot be all programmed by human -- it must have its own ability to learn, adapt and evolve. On the other hand, researches in machine learning have created powerful algorithms to find the underlying rules in the data and make accurate predictions based on the learned rules. These algorithms could be seen as an improved version of traditional statistics, being able to work in a much larger scale. They can find inherent nonlinear relations among hundreds and thousands of variables, with up to millions of observations. Nowadays, massive amounts of data are generated from every corner of the world in a daily basis. Therefore, machine learning, an automated way to reason and make sense from these data, could only become more and more important. In this talk, I will discuss various aspects of machine learning: the intuition, the paradigms, important algorithms and applications and limitations.
机器学习研究的问题是如何从现实数据中获取有用的信息及规则,并利用这些规则做出预测。在人工智能中,这一问题极为重要。因为一台完全按照人编好的规则进行运算的计算机,不能称之为一个真正的智能体。一个真正的智能体,必须有自己从外部世界的数据中归纳,学习,适应和演化的能力。近年来,在机器学习的研究中,产生了一些十分强大的算法,可以在数据中总结归纳隐藏的规律并据此做出预测和判断。由于其方法的广泛应用,机器学习已经几乎摆脱人工智能而成为了一个独立的学科。从数据分析的角度来看,机器学习方法可认为是传统统计学的“加强版”,它能够在成千上万个变量和数百万的观测数据中,寻找潜在的复杂非线性规律。当今世界,在各行各业中,观测数据的量都迅速增长,远远超过人力所能分析的范围。机器学习提供了自动分析这些数据的方法,因此,其在各个科学工程领域的重要性与日俱增。在本次报告中,我将着重介绍机器学习的基本理念、基本方法,应用范围,并举出一些在实际问题中应用机器学习的实例,希望这些新的数据分析工具能在更多的科学技术问题中得到更好的应用。
报告人简介:
Dr. Fuxin Li received his B.E. in Computer Science and Engineering from Zhejiang University in 2001. He studied Pattern Recognition and Intelligent Systems in the Institute of Automation, Chinese Academy of Science from 2003 to 2008, where he received a Ph.D. In 2009. Starting from September 2008, he works at the Institute of Numerical Simulation, University of Bonn for a post-doctoral position. His current research direction is machine learning and computer vision. He has published over 10 papers in machine learning and proteomics.
李伏欣博士2001年在浙江大学获得学士学位。2003-2008年,他在中科院自动化所攻读博士学位,专业为模式识别与智能系统。他于2009获得博士学位。2008年9月起,他在波恩大学数值模拟研究所做博士后。目前,他的研究方向是机器学习与计算机视觉。他已在机器学习与蛋白质组学方向上发表了10余篇论文。
荣誉:
2005年“微软学者”奖学金
“Microsoft Fellow” Award in 2005