@article{oai:bunkyo.repo.nii.ac.jp:00003132,
 author = {鈴木, 昇一},
 journal = {情報研究, Information and Communication Studies},
 month = {2005-07-01, 2010-02-03},
 note = {It is necessary to measure a degree SM of similarity between an input pattern and each of memorized representative patterns.SM constructed strictly and minutely here has an integral nuclear function, which is different from traditional similarities. We costruct four kinds of SM s which must satisfy three properies of an orthonomality, a nomalization and an invariance under a modelconstruction operator, from points of view of an inner product, a gaussian probability density, a categorical a-posteriori probability and a least square criterion. SM s presented here are useful for recognizing patterns that are complicated in structure, which are of value for reference in designing similarity-measure functions which are able to be used to good effect for patterns in question to be recognized.},
 pages = {111--147},
 title = {正規直交性を満たす類似度関数SMに積分核が存在するか?},
 volume = {33},
 year = {}
}