2.txt

This complete matlab for neural network

发信人: liumangtu (流氓兔), 信区: DataMining
标  题: 18号声西楼203#，E.D.Goodman的学术报告
发信站: 南京大学小百合站 (Tue Dec 17 11:08:34 2002)

报告的主题是：
Automated Design of Multi-Domain Dynamic Systems Using Evolutionary Algorithms


ABSTRACT: Genetic algorithms are widely applied for optimization of parameters
 for systems of fixed topology.  Genetic programming is widely used to synthes
ize tree-like structures by interconnecting basic "atomic" components.  Howeve
r, automated synthesis of complex dynamic systems (control systems, etc.) invo
lving electrical, mechanical, hydraulic, pneumatic, and/or thermal components 
is difficult -- the synthesis system must simultaneously generate and optimize
 structures and associated parameter values, and the modeling/analysis/simulat
ion system must be able to handle all of these energy domains simultaneously. 
    

 The bond graph/genetic programming (BG/GP) system is being developed to addre
ss such problems.  Bond graphs are used to capture the power and energy relati
onships among components from arbitrary energy domains, allowing formulation o
f state equations describing the system's dynamics. The user specifies an "emb
ryo" system that includes the input/output where performance objectives will b
e measured, and the system produces a sequence of designs (for example, combin
ations of mechanical and analog electronic components) that become increasingl
y good, for many sample applications.  Issues on which significant advances ha
ve been made include 1) Balancing of the search between finding new structures
 and finding appropriate parameters for them. 2) Dramatically improving the sy
stem's capability to search effectively, without premature convergence, and to
 combine good "building blocks" to generate still better ones. 3) Finding appr
opriate primitive elements from which to construct the bond graphs more effect
ively for a given problem domain.

主办：南京大学电子科学与工程系
协办：江苏省电子学会天线与微波专业组，IEEE AP-MTT-EMC Nanjing Chapter
华东师范大学技术学院
时间：2002年12月18日 2:00-5:00pm  
地点：南京大学 声西(声学楼西侧) 203 

--

※ 来源:．南京大学小百合站 http://bbs.nju.edu.cn [FROM: 202.119.44.21]