2026-03-09T23:55:30Z https://bunkyo.repo.nii.ac.jp/oai

oai:bunkyo.repo.nii.ac.jp:00003387 2024-08-08T01:03:53Z 1:26:195

キゾク ケイスウホウ ニ モトズク ルイジド キゾク カンケイ アイマイド ニンシキ ジョウホウリョウ ノ ケイサンキ シミュレーション 帰属係数法に基づく類似度,帰属関係あいまい度,認識情報量の計算機シミュレーション Similarity, Membership Equivocation and Amount of Recognitive Information Relating to Membership Coefficient Method and its Computer Simulation 鈴木, 昇一 We point out three following matters : ( 1 ) It does not necessarily follow that a similarity-measure is defined between two any patterns. ( 2 ) The measure have only to be defined between any pattern and a typical pattern of any category. ( 3 ) The measure have to be defined using a perceptual representation of any pattern, that is, a model of any pattern. 　It is the purpose of the present paper to propose by the use of membership-coefficients the measure which satisfies three above matters and to compute by means of a computer simulation membership-equivocations and amounts of recognitive information of isolated Japanese vowels as input patterns. The equivocation is defined as an entropy function of the measure, and the amount as a Kullback discriminant function of the measure. 　We pay attention to the following request : An amount of selfinformation must be obtained as a logarithmic function of the reciprocal of an occurrence-probability of a category if a system recognizes that an input pattern belongs to the category. The request can be explained sufficiently with such an amount of recognitive information. departmental bulletin paper 文教大学 1990-01-01 VoR application/pdf 情報研究 11 51 68 Information and Communication Studies 03893367 https://bunkyo.repo.nii.ac.jp/record/3387/files/BKSJ110005.pdf https://bunkyo.repo.nii.ac.jp/records/3387 jpn