The Seventeenth Computational Anatomy Seminar

Overview

•Date & Time ◦July 18 16:35 - 17:35, July19 10:40 - 11:40, 12:50-13:50, 2013

•Venue Tohoku University, Katahira campus, Sakura hall

•Admission ◦Free

Program

  • Lecture 1 : July 18 2013 16:35 - 17:35
    • Makoto Yoshizawa(Tohoku University)
    • Title
      • Cyber-Medicine −From VR-Rehabilitation System to Electronic Doctor’s Bag for Disaster Areas−
    • Abstract
      • On March 11, the giant tsunami in the Great East Japan Earthquake hit the coastal areas of Tohoku district and fully destroyed their environment of life. However, the healthcare environment in Tohoku region has already been in a critical state because the uneven distribution of medical doctors has been widened due to an aging and declining population. It can be easily predicted that the disaster will accelerate the deterioration of this serious problem. In this situation, we developed a prototype of communications system, named “Electronic Doctor’s Bag” in 2009 to provide the ubiquitous communications system not only for home-visit medical services but also for mass health examination, emergency care, and disaster areas. In the present lecture, it will report the results of the field tests of the proposed telemedical tool applied in an isolated island and a disaster area of the Great East Japan Earthquake, introducing other high-tech ICT systems for rehabilitation and health care.

  • Lecture 2 : July 19 2013 10:40 - 11:40
    • Hiroshi Matsuzoe (Nagoya Institute of Technology )
    • Title
      • Geometry of Statistical Models from the Viewpoint of Differential Geometry
    • Abstract
      • In geometric theory of statistical inferences, geometry of parameter spaces of probability distributions is called information geometry. In information geometry, a parametric statistical model is regarded as a Riemannian manifold, and a pair of dualistic affine connections plays an important role. An exponential family is a typical statistical model in statistics, which includes the set of Gaussian distributions. An exponential family has Hessian structures, and these geometric quantities are induced form the Kullback-Leibler divergence. On the other hand, probability densities of non-exponential type are useful in the theory of complex systems. A deformed exponential family is an example of a set of such probability distributions, and is introduced as a generalization of exponential families. A deformed exponential family naturally has two kinds of Hessian structures. In this presentation, we discuss these geometric structures and explain relations to geometry of divergences. In addition, from discussions above, we find that notions of expectations and independences for random variables are dependent to the given statistical model. As applications, we consider a generalization of independences and generalize the maximum likelihood methods.

  • Lecture 3 : July 19 2013 12:50-13:50
    • Takayuki Okatani(Tohoku University)
    • Title
      • Learning Features by Deep Networks and Its Application to Image Recognition
    • Abstract
      • Deep learning is a methodology in machine learning that uses multi-layered neural networks. It have gained a lot of attention lately since it showed considerably higher performance than previous methods when being applied to problems in different areas such as image recognition and speech recognition. In this talk, I will present an overview of a collection of the methods that have been known so far with a focus on applications to image recognition and will try to sort out them.


Inquiry

  • Akinobu Shimizu (Tokyo Univ. of Agriculture and Technology)
    • E-mail simiz @ cc.tuat.ac.jp

トップ   編集 差分 バックアップ 添付 複製 名前変更 リロード   新規 一覧 単語検索 最終更新   ヘルプ   最終更新のRSS
Last-modified: 2013-11-11 (Mon) 09:45:08 (1438d)