WEKO3
インデックスリンク
アイテム
{"_buckets": {"deposit": "354b352e-3999-49e5-ba6a-07d0657736df"}, "_deposit": {"created_by": 3, "id": "3243", "owners": [3], "pid": {"revision_id": 0, "type": "depid", "value": "3243"}, "status": "published"}, "_oai": {"id": "oai:bunkyo.repo.nii.ac.jp:00003243", "sets": ["230"]}, "author_link": ["4198"], "item_5_alternative_title_1": {"attribute_name": "タイトル カナ", "attribute_value_mlt": [{"subitem_alternative_title": "2カテゴリ ブンルイ コンナンド ノ ジョウホウ リロン"}]}, "item_5_biblio_info_13": {"attribute_name": "書誌情報", "attribute_value_mlt": [{"bibliographicIssueDates": {"bibliographicIssueDate": "2001-01-01", "bibliographicIssueDateType": "Issued"}, "bibliographicPageEnd": "160", "bibliographicPageStart": "63", "bibliographicVolumeNumber": "26", "bibliographic_titles": [{"bibliographic_title": "情報研究"}, {"bibliographic_title": "Information and Communication Studies", "bibliographic_titleLang": "en"}]}]}, "item_5_date_43": {"attribute_name": "作成日", "attribute_value_mlt": [{"subitem_date_issued_datetime": "2011-02-23", "subitem_date_issued_type": "Created"}]}, "item_5_description_12": {"attribute_name": "抄録", "attribute_value_mlt": [{"subitem_description": " パターン認識の数学的理論(SS理論)では、入力パターンφに対応するパターンモデルTφを求め、Tφから不動点パターンモデルを連想する形で、φの帰属するカテゴリを決定する多段階パターン変換連想形不動点認識法(SS連想形不動点認識法)が考えられている。\n このパターン認識法を採用している本研究では、各出力出現確率q(Tφ)(q∈Φ[t1,t2])と、出力Tφが観測された条件の下で各入力?n[j]の再現確率p(?n[j]/Tφ)(n∈{1,2})とを与え、\n AMI(?[j];Φ[t1,t2])\n =2Σn=1 Σφ∈Φ[t1,t2] \n q(Tφ)・p(?n[j]/Tφ)・\n loge[p(?n[j]/Tφ)/p(?n[j])]\nの最大値を求めることに関連し、2カテゴリ分類困難度DOC(φ,?[j];SM),2カテゴリ分類容易度EOC(φ,?[j];SM),曖昧度H(?[j]/Tφ)に関する解析が展開される。\n 曖昧度H(?[j]/Tφ)の、p(?1[j]/Tφ)に関する微係数を2カテゴリ分類困難度DOC(φ,?[j];SM)と定義することから、解析が始まっているが、SS理論での不動点多段階想起形認識[B3],[B4]は解消される不確定さAMI(?[j];Φ[t1,t2])が大きくなるようなパターン処理法であるとの結論が本研究によって鮮明にされる。\n 本研究は、SS理論のaxiom 2を満たす類似度関数SMを式(3.63)のように事後確率p(?1[j]/Tφ)と設定している故に(この設定は本研究独創性を確実なものにしている)、得られた研究内容は設定された1つの認識の働きが処理の対象とする問題のパターンφの集合Φ[t1,t2]に適切なものであるかどうかを検証する場面で決定的な役割を果たすという意味で、認識システムRECOGNITRON[B3],t[B4]の構成に信頼を与えるものである。\n\\n A recognition system RECOGNITRON which has been presented in a mathematical theory (i.e.SS theory) of recognizing patterns suggested by S.Suzuki gets a corresponding pattern-model Tφ of an input pattern φ in question to be recognized, and determines a category to which φ belongs so that a fixedpoint pattern-model that appeared on a final stage of a muti-stage structural-fertilization transformation of pattern-models may be recalled in such a way of solving a fixed-point equation of associative reconition about Tφ.\n In this recognition method an analysis about an difficulty DOC(φ,?[j];SM) and an easiness EOC(?[j];SM) of binary classification, and an equivocation H(?[j]/Tφ) is developed in full seeking for the maximum of an average amount\n AMI(?[j];Φ[t1,t2])\n =2Σn=1 Σφ∈Φ[t1,t2]\n q(Tφ)・p(?n[j]/Tφ)・\n loge[p(?n[j]/Tφ)/p(?n[j])]\n of mutual information, where\n q(Tφ(φ∈Φ[t1,t2]) is a probability of occurrences of the pattern-model Tφ, and p(?n[j]/Tφ) (n∈{1,2}) is a conditional probability of occurrences of the n-th category ?n[j] given that Tφ has occurred at some trial.\n The above-mentioned analysis begins with defining DOC(φ,?[j];SM) as an differential value of the function H(?[j]/Tφ) concerning the variable p(?1[j]/Tφ). We can conclude that the recognition method proposed by S.Suzuki maximizes AMI(?[j];Φ[t1,t2]) which is an amount of un ertainty of ?[j]≡{?1[j],?2[j]} removed after many observations of Tφ.\n In the above analysis we adopt the similarity-measure SM satisfying axiom 2 as the aposteriori probability p(?1[j]/Tφ), which makes sure of an originality of this study.\n The obtained result can play a definite part in verifying whether or not a selected recognition method is suitable for the set Φ[t1,t2] of patterns in question, which therefore gives a reliability to a construction of the recognition system RECOGNITRON [B3],[B4].", "subitem_description_type": "Abstract"}]}, "item_5_description_34": {"attribute_name": "資源タイプ", "attribute_value_mlt": [{"subitem_description": "text", "subitem_description_type": "Other"}]}, "item_5_description_38": {"attribute_name": "フォーマット", "attribute_value_mlt": [{"subitem_description": "application/pdf", "subitem_description_type": "Other"}]}, "item_5_publisher_16": {"attribute_name": "出版者", "attribute_value_mlt": [{"subitem_publisher": "文教大学"}]}, "item_5_source_id_19": {"attribute_name": "ISSN", "attribute_value_mlt": [{"subitem_source_identifier": "03893367", "subitem_source_identifier_type": "ISSN"}]}, "item_5_text_14": {"attribute_name": "出版年", "attribute_value_mlt": [{"subitem_text_value": "2001"}]}, "item_5_text_15": {"attribute_name": "出版地", "attribute_value_mlt": [{"subitem_text_value": "神奈川県茅ヶ崎市"}]}, "item_5_text_17": {"attribute_name": "出版社 カナ", "attribute_value_mlt": [{"subitem_text_value": "ブンキョウ ダイガク"}]}, "item_5_text_18": {"attribute_name": "Publisher", "attribute_value_mlt": [{"subitem_text_value": "Bunkyo University"}]}, "item_5_text_39": {"attribute_name": "本文言語", "attribute_value_mlt": [{"subitem_text_value": "日本語"}]}, "item_5_text_42": {"attribute_name": "ID", "attribute_value_mlt": [{"subitem_text_value": "BKSJ260003"}]}, "item_5_text_45": {"attribute_name": "更新日(XooNIps)", "attribute_value_mlt": [{"subitem_text_value": "2011/6/7"}]}, "item_5_text_46": {"attribute_name": "記録日", "attribute_value_mlt": [{"subitem_text_value": "2011/2/23"}]}, "item_5_text_47": {"attribute_name": "最終更新日(XooNIps)", "attribute_value_mlt": [{"subitem_text_value": "2011/6/7"}]}, "item_5_text_48": {"attribute_name": "更新履歴(XooNIps)", "attribute_value_mlt": [{"subitem_text_value": "Jun 7, 2011 更新日, 抄録 を変更"}, {"subitem_text_value": "May 11, 2011 インデックス, 更新日 を変更"}, {"subitem_text_value": "Feb 23, 2011 フリーキーワード, インデックス, 抄録, キーワード を変更"}]}, "item_5_text_49": {"attribute_name": "登録者(XooNIps)", "attribute_value_mlt": [{"subitem_text_value": "bunkyo2"}]}, "item_5_text_50": {"attribute_name": "閲覧数(XooNIps)", "attribute_value_mlt": [{"subitem_text_value": "602"}]}, "item_5_text_51": {"attribute_name": "ダウンロード数(XooNIps)", "attribute_value_mlt": [{"subitem_text_value": "612"}]}, "item_5_text_7": {"attribute_name": "Author", "attribute_value_mlt": [{"subitem_text_value": "Suzuki, Shoichi"}]}, "item_5_text_8": {"attribute_name": "所属機関", "attribute_value_mlt": [{"subitem_text_value": "文教大学情報学部"}]}, "item_5_text_9": {"attribute_name": "Institution", "attribute_value_mlt": [{"subitem_text_value": "Bunkyo University Faculty of Information and Communications"}]}, "item_5_version_type_35": {"attribute_name": "著者版フラグ", "attribute_value_mlt": [{"subitem_version_resource": "http://purl.org/coar/version/c_970fb48d4fbd8a85", "subitem_version_type": "VoR"}]}, "item_creator": {"attribute_name": "著者", "attribute_type": "creator", "attribute_value_mlt": [{"creatorNames": [{"creatorName": "鈴木, 昇一"}], "nameIdentifiers": [{"nameIdentifier": "4198", "nameIdentifierScheme": "WEKO"}]}]}, "item_files": {"attribute_name": "ファイル情報", "attribute_type": "file", "attribute_value_mlt": [{"accessrole": "open_date", "date": [{"dateType": "Available", "dateValue": "2018-03-24"}], "displaytype": "detail", "download_preview_message": "", "file_order": 0, "filename": "BKSJ260003.pdf", "filesize": [{"value": "5.0 MB"}], "format": "application/pdf", "future_date_message": "", "is_thumbnail": false, "licensetype": "license_free", "mimetype": "application/pdf", "size": 5000000.0, "url": {"label": "BKSJ260003.pdf", "url": "https://bunkyo.repo.nii.ac.jp/record/3243/files/BKSJ260003.pdf"}, "version_id": "59bc204e-d2e5-4102-a6d0-10d6cf1cc5a1"}]}, "item_keyword": {"attribute_name": "キーワード", "attribute_value_mlt": [{"subitem_subject": "パターン認識の数学的理論(SS理論)", "subitem_subject_scheme": "Other"}, {"subitem_subject": "モデル構成作用素", "subitem_subject_scheme": "Other"}, {"subitem_subject": "類似度関数", "subitem_subject_scheme": "Other"}, {"subitem_subject": "大分類関数", "subitem_subject_scheme": "Other"}, {"subitem_subject": "不動点連想形多段階認識法", "subitem_subject_scheme": "Other"}, {"subitem_subject": "曖昧度", "subitem_subject_scheme": "Other"}, {"subitem_subject": "平均相互情報量", "subitem_subject_scheme": "Other"}, {"subitem_subject": "シグモイド関数", "subitem_subject_scheme": "Other"}, {"subitem_subject": "2カテゴリ分類困難度・容易度", "subitem_subject_scheme": "Other"}, {"subitem_subject": "最大類似度認識法", "subitem_subject_scheme": "Other"}]}, "item_language": {"attribute_name": "言語", "attribute_value_mlt": [{"subitem_language": "jpn"}]}, "item_resource_type": {"attribute_name": "資源タイプ", "attribute_value_mlt": [{"resourcetype": "departmental bulletin paper", "resourceuri": "http://purl.org/coar/resource_type/c_6501"}]}, "item_title": "2カテゴリ分類困難度の情報理論", "item_titles": {"attribute_name": "タイトル", "attribute_value_mlt": [{"subitem_title": "2カテゴリ分類困難度の情報理論"}, {"subitem_title": "Information Theory of Difficulty of Two-Category Classification", "subitem_title_language": "en"}]}, "item_type_id": "5", "owner": "3", "path": ["230"], "permalink_uri": "https://bunkyo.repo.nii.ac.jp/records/3243", "pubdate": {"attribute_name": "公開日", "attribute_value": "2011-02-23"}, "publish_date": "2011-02-23", "publish_status": "0", "recid": "3243", "relation": {}, "relation_version_is_last": true, "title": ["2カテゴリ分類困難度の情報理論"], "weko_shared_id": -1}
2カテゴリ分類困難度の情報理論
https://bunkyo.repo.nii.ac.jp/records/3243
https://bunkyo.repo.nii.ac.jp/records/3243e1345b93-8abe-4daf-adf0-2eb4b9d78724
名前 / ファイル | ライセンス | アクション |
---|---|---|
BKSJ260003.pdf (5.0 MB)
|
|
Item type | 紀要論文 / Departmental Bulletin Paper(1) | |||||
---|---|---|---|---|---|---|
公開日 | 2011-02-23 | |||||
タイトル | ||||||
タイトル | 2カテゴリ分類困難度の情報理論 | |||||
タイトル | ||||||
タイトル | Information Theory of Difficulty of Two-Category Classification | |||||
言語 | ||||||
言語 | jpn | |||||
キーワード | ||||||
主題 | パターン認識の数学的理論(SS理論) | |||||
キーワード | ||||||
主題 | モデル構成作用素 | |||||
キーワード | ||||||
主題 | 類似度関数 | |||||
キーワード | ||||||
主題 | 大分類関数 | |||||
キーワード | ||||||
主題 | 不動点連想形多段階認識法 | |||||
キーワード | ||||||
主題 | 曖昧度 | |||||
キーワード | ||||||
主題 | 平均相互情報量 | |||||
キーワード | ||||||
主題 | シグモイド関数 | |||||
キーワード | ||||||
主題 | 2カテゴリ分類困難度・容易度 | |||||
キーワード | ||||||
主題 | 最大類似度認識法 | |||||
資源タイプ | ||||||
資源タイプ識別子 | http://purl.org/coar/resource_type/c_6501 | |||||
資源タイプ | departmental bulletin paper | |||||
著者 |
鈴木, 昇一
× 鈴木, 昇一 |
|||||
著者 | ||||||
Suzuki, Shoichi | ||||||
所属機関 | ||||||
文教大学情報学部 | ||||||
所属機関 | ||||||
Bunkyo University Faculty of Information and Communications | ||||||
内容記述 | ||||||
内容記述タイプ | Abstract | |||||
内容記述 | パターン認識の数学的理論(SS理論)では、入力パターンφに対応するパターンモデルTφを求め、Tφから不動点パターンモデルを連想する形で、φの帰属するカテゴリを決定する多段階パターン変換連想形不動点認識法(SS連想形不動点認識法)が考えられている。 このパターン認識法を採用している本研究では、各出力出現確率q(Tφ)(q∈Φ[t1,t2])と、出力Tφが観測された条件の下で各入力?n[j]の再現確率p(?n[j]/Tφ)(n∈{1,2})とを与え、 AMI(?[j];Φ[t1,t2]) =2Σn=1 Σφ∈Φ[t1,t2] q(Tφ)・p(?n[j]/Tφ)・ loge[p(?n[j]/Tφ)/p(?n[j])] の最大値を求めることに関連し、2カテゴリ分類困難度DOC(φ,?[j];SM),2カテゴリ分類容易度EOC(φ,?[j];SM),曖昧度H(?[j]/Tφ)に関する解析が展開される。 曖昧度H(?[j]/Tφ)の、p(?1[j]/Tφ)に関する微係数を2カテゴリ分類困難度DOC(φ,?[j];SM)と定義することから、解析が始まっているが、SS理論での不動点多段階想起形認識[B3],[B4]は解消される不確定さAMI(?[j];Φ[t1,t2])が大きくなるようなパターン処理法であるとの結論が本研究によって鮮明にされる。 本研究は、SS理論のaxiom 2を満たす類似度関数SMを式(3.63)のように事後確率p(?1[j]/Tφ)と設定している故に(この設定は本研究独創性を確実なものにしている)、得られた研究内容は設定された1つの認識の働きが処理の対象とする問題のパターンφの集合Φ[t1,t2]に適切なものであるかどうかを検証する場面で決定的な役割を果たすという意味で、認識システムRECOGNITRON[B3],t[B4]の構成に信頼を与えるものである。 \n A recognition system RECOGNITRON which has been presented in a mathematical theory (i.e.SS theory) of recognizing patterns suggested by S.Suzuki gets a corresponding pattern-model Tφ of an input pattern φ in question to be recognized, and determines a category to which φ belongs so that a fixedpoint pattern-model that appeared on a final stage of a muti-stage structural-fertilization transformation of pattern-models may be recalled in such a way of solving a fixed-point equation of associative reconition about Tφ. In this recognition method an analysis about an difficulty DOC(φ,?[j];SM) and an easiness EOC(?[j];SM) of binary classification, and an equivocation H(?[j]/Tφ) is developed in full seeking for the maximum of an average amount AMI(?[j];Φ[t1,t2]) =2Σn=1 Σφ∈Φ[t1,t2] q(Tφ)・p(?n[j]/Tφ)・ loge[p(?n[j]/Tφ)/p(?n[j])] of mutual information, where q(Tφ(φ∈Φ[t1,t2]) is a probability of occurrences of the pattern-model Tφ, and p(?n[j]/Tφ) (n∈{1,2}) is a conditional probability of occurrences of the n-th category ?n[j] given that Tφ has occurred at some trial. The above-mentioned analysis begins with defining DOC(φ,?[j];SM) as an differential value of the function H(?[j]/Tφ) concerning the variable p(?1[j]/Tφ). We can conclude that the recognition method proposed by S.Suzuki maximizes AMI(?[j];Φ[t1,t2]) which is an amount of un ertainty of ?[j]≡{?1[j],?2[j]} removed after many observations of Tφ. In the above analysis we adopt the similarity-measure SM satisfying axiom 2 as the aposteriori probability p(?1[j]/Tφ), which makes sure of an originality of this study. The obtained result can play a definite part in verifying whether or not a selected recognition method is suitable for the set Φ[t1,t2] of patterns in question, which therefore gives a reliability to a construction of the recognition system RECOGNITRON [B3],[B4]. |
|||||
書誌情報 |
情報研究 en : Information and Communication Studies 巻 26, p. 63-160, 発行日 2001-01-01 |
|||||
ISSN | ||||||
収録物識別子 | 03893367 | |||||
著者版フラグ | ||||||
出版タイプ | VoR | |||||
フォーマット | ||||||
内容記述タイプ | Other | |||||
内容記述 | application/pdf | |||||
本文言語 | ||||||
日本語 | ||||||
ID | ||||||
BKSJ260003 | ||||||
作成日 | ||||||
日付 | 2011-02-23 |