Home Sitemap Contact 中文 CAS
  • HOME
  • About Us
  • Research
  • People
  • International Cooperation
  • News
  • Education & Training
  • Join Us
  • Publications
  • Papers
  • Resources
  • Life at ICT
  • Links
  • Location:Home>News>Upcoming Events
    Ensembles of Intelligent Systems with Generalization Capabilities for Complex Application Problems
    Update time: 2011-06-22
    Text Size: A A A

    Speaker: Prof. Yi Lu, University of Michigan-Dearborn

    Time: 14:00 pm – 16:00 pm, June 22th, 2011 (Wednesday)

    Place: Auditorium in 4th Floor, ICT, CAS


    Many real world problems are too large and complex for a single system to solve. The idea of using an ensemble of intelligent systems to solve a large and complex problem is consistent with many natural intelligent systems: solve a complex problem by decomposing it into smaller problems that are being solved by individual subsystems. A complex problem usually implies that the underlying system, whether it is a physical, biological, or social system, contains a large number of diverse, dynamic and interdependent components, has no definitive problem boundary, demands high accuracy in a testable solution, and requires the solution to generalize to a broad range of similar systems. Examples of complex problems include ant colonies, human economies, climate, nervous systems, modern energy or telecommunication infrastructures, and engineering fault diagnostics. 

    In this talk, I will discuss our research in ensemble learning with respect to three generalization issues that are particularly important for solving complex problems: an ensemble trained on one data collection can generalize to other data collections acquired from similar but different physical models; an ensemble trained on a suite of large data collections acquired from multiple physical models can generalize to the unseen data acquired from similar physical models; and ensembles that have the capability of incremental learning and generalization. I will present a two-step ensemble approach for ensemble learning, and algorithms for ensemble selection, and trainable ensemble function. I wil show through experiments that a neural network ensemble designed and trained by the proposed methodology, the proposed ensemble selection and decision functions can generalize well in solving a complex problem of onboard vehicle fault diagnostics.


    Prof. Yi Lu Murphey received a M.S. degree in computer science from Wayne State University, Detroit, Michigan, in 1983, and a Ph.D degree with a major in Computer Engineering and a minor in Control Engineering from the University of Michigan, Ann Arbor, Michigan, in 1989. From 1989 to 1992, she was a research scientist at the Environmental Research Institute of Michigan, Ann Arbor, Michigan. She joined the faculty of the ECE department at the University of Michigan-Dearborn in 1992. Currently she is a professor and the Chair of the Electrical and Computer Engineering Department at the University of Michigan-Dearborn. Professor Murphey is actively involved in funded research in the areas of machine learning, computer vision and intelligent systems with applications to engineering diagnostics, optimal vehicle power management, text data mining, and robotic vehicles. More details of her research can be found at http://www-personal.engin.umd.umich.edu/~yilu/ .

    Address :No.6 Kexueyuan South Road Zhongguancun,Haidian District Beijing,China
    Postcode :100190 Tel : (8610)62601166 Email : office@ict.ac.cn