著名专家J.J.HOPFIELD教授在南航讲学

世界著名神经元网络专家、神经元网络创始人之一,美国科学院士、加州理工学院计算与神经系主任、化学及生物学教授John J. Hopfield博士,应我国863高技术计划智能计算机专家组和南京航空学院的联合邀请,在南航作了为期2.5周“人工神经网络”讲座。Hopfield教授系统地讲授了神经网络基本原     

A review of: “FUZZY SETS,UNCERTAINTY, AND INFORMATION,” by George J. Klir and Tina A. Folger. Prentice Hall,Englewood Cliffs,N.J., 1988, XI+ 355 pp.

Possibility distribution introduced by Zadeh ["Fuzzy sets as a basis for a theory of possibility theory", Fuzzy Sets Syst. 1 (1978) 3-28] in his introductory paper of possibility theory assumes a normal distribution, in the sense that it supposes the existence of at least one element s 0 of the universe of discourse U , for which the distribution ~ is fully compatible with the context of interest: ~ (s 0 )=1. However, when such element does no longer exist, it leads to a subnormal possibility distribution. This situation may arise from incomplete data, inconsistent statements, or contradictory beliefs. To deal with such case, many authors like Yager ["On the relationships of methods of aggregation evidence in expert systems", Cybern. Syst. , 16 (1985) 1-21; "A modification of the certainty measure to handle subnormal distributions", Fuzzy Sets Syst. , 20 (1986) 317-324], Dubois and Prade ["An alternative approach to the handling of subnormal possibility distributions--A critical comment on a proposal of Yager", Fuzzy Sets Syst. , 24 (1987) 123-126] have put forward some proposals in order to keep track of the consistency of the basic axioms attached to possibility and necessity measures. In this paper, the proposals are reviewed in the light of new results regarding some appealing criteria. Particularly, when subnormal distribution and normal distribution are encountered in the same level, intuitively, two approaches are possible: Either the subnormal distributions are risen up to a normal distribution level, or the normal ones are flatted down to agree with the normal ones. In both cases there is a sort of gaining or losing information. We review some of the proposal solutions. The flatting approach is mainly related to fuzzy arithmetic calculus while the rising effect is motivated by Dempster-Shafer theory of evidence and its normalization paradigm. The two approaches will also be investigated with respect to some appealing criteria like preference preservation, distance minimization, entropy, minimum/maximum specificity, and, further, particular interest is focused on information based uncertainty preservation. Later on, the proposals are discussed according to the f -certainty qualification where the greatest value h of the subnormal distribution is understood as a degree of certainty that must be attached to the resulting normal distribution.     

Michael J. Miller 随身携带

你喜欢随身携带哪种机器?这一问题随着新的轻便型个人电脑在今年进入市场而变得更加复杂了。对于那些带着便携机经常四处奔波的人来说,更小、更轻的设备——即我们曾经说过的个人数字助理(PDAs),其历史可以追溯到Apple的Newton造成巨大轰动的年代——是一种奢望。但当时那些设备的用途与现在的差别甚大,而且只能提供很少