特徴空間の動的構成によるプローブデータのリアルタイム補完技術
8
0
0
全文
(2) (a) FCD history. Feature space analysis feature bases. (b) Current FCD. Weighted projection into the feature space. Projection point (c) Estimation data. Inverse projection into the original space Feature space. Fig. 1 Process of the realtime imputation. 䊨䊷䊑䊂䊷䉺䉕 ⴕ 䈱〝 䉶䊮䉰䈫ห ╬ 䈱 5 ಽ ᦼ 䈱䊂䊷䉺䈫䈚䈩 ↪ 䈜䉎႐ ว 䇮ห ᤨ ೞ 䈪䈱䊂䊷䉺䈱ᰳ ៊ ₸ 䈲ో 䈱 9 ഀ 䈮㆐ 䈜䉎䇯 䉋䈦䈩〝 䉶䊮䉰䈫ห ᭽ 䈮ᛒ 䈉႐ ว 䇮ᰳ ៊ 䈚 䈩䈇䉎䊂䊷䉺䈱 ቢ ᚻ Ბ 䈏ᔅ ⷐ 䈮䈭䉎䇯 ቢ ᚻ Ბ 䈱৻ ⥸ ⊛ 䈭ᚻ ᴺ 䈫䈚䈩䇮ㆊ 䈱 䊒䊨䊷䊑䊂䊷䉺䈱ห ᤨ ೞ ᐔ ဋ ୯ 䉕 ቢ 䊂䊷 䉺䈫䈚䈩ឭ ଏ 䈜䉎ᚻ ᴺ 䈏䈅䉎䇯䈚䈎䈚䇮䈖䈱ᚻ ᴺ 䈲 ቯ 䈚䈢 ቢ ᖱ ႎ 䉕ឭ ଏ 䈜䉎䈖䈫䈲䈪䈐 䉎䈏䇮ᦐ ᣣ 䉇ቄ ▵ 䈱 ᄌ ൻ 䈮 ච ಽ 䈮 ኻ ᔕ 䈪䈐 䈭䈇䇯䉁䈢ㆊ 䊂䊷䉺䉕䇮ᦐ ᣣ 䇮ቄ ▵ 䈱䉋䈉䈮 ⚦ 䈮ಽ 㘃 䈚䈩䇮䈠䉏䈡䉏䈮䈧䈇䈩ห ᤨ ೞ ᐔ ဋ ୯ 䉕᳞ 䉄䉎ᚻ ᴺ 䉅⠨ 䈋䉌䉏䉎䈏䇮ಽ 㘃 න 䈗䈫䈮䉰䊮䊒䊦ᢙ 䈏ዋ 䈭䈒䈭䉍䇮⛔ ⸘ ⊛ 䈭 ା 㗬 ᕈ 䈲ૐ ਅ 䈜䉎䇯 䈖䈱⸃ ╷ 䈫䈚䈩䇮․ ᓽ ⓨ 㑆 䉕↪ 䈇䈢䊒䊨 䊷䊑䉦䊷䊂䊷䉺䈱䊥䉝䊦䉺䉟䊛 ቢ ᛛ ⴚ 䈏ႎ ๔ 䈘䉏䈩䈇䉎[2]䇯䈖䈱ᛛ ⴚ 䈲䇮ㆊ 䈱䊒䊨䊷 䊑䊂䊷䉺䈎䉌․ ᓽ ⓨ 㑆 䉕↢ ᚑ 䈚䇮 ᴫ 䈱䊒䊨 䊷䊑䊂䊷䉺䉕䈠䈱ᰳ ៊ 䈮ᔕ 䈛䈩․ ᓽ ⓨ 㑆 䈮 ᓇ 䈜䉎䈖䈫䈪䇮ᰳ ៊ ୯ 䈱 ቢ 䉕ⴕ 䈉䇯䈖䈱․ ᓽ ⓨ 㑆 䈲 〝 䊥䊮䉪㑆 䈱 ㅢ ᖱ ႎ 䈱⋧ 㑐 㑐 ଥ 䉕 䈚䈩䈇䉎䇯 䈖䈱䊥䉝䊦䉺䉟䊛 ቢ ᛛ ⴚ 䈪䈲䇮䊒䊨䊷䊑䉦. 䊷䈱䊥䊮䉪䉦䊋䊷₸ 䋨 ో 䊥䊮䉪ᢙ 䈮ኻ 䈜䉎䊒䊨 䊷䊑 ㅢ ᖱ ႎ 䈏 㓸 䈪䈐䈢䊥䊮䉪ᢙ 䈱ഀ ว 䋩 䈏 20%⒟ ᐲ 䈅䉏䈳䇮 ല 䈭 ቢ 䉕ⴕ 䈉䈖䈫䈏 䈪䈐䈢䇯䈚䈎䈚䇮 ᣇ ㇺ Ꮢ 䈭䈬䊒䊨䊷䊑ᖱ ႎ 䈱 㓸 䈏࿎ 㔍 䈭 ၞ 䈪䈲䇮䊥䊮䉪䉦䊋䊷₸ 䋲䋰䋦 䈱㆐ ᚑ 䉅ኈ ᤃ 䈪䈲䈭䈇䇯䈖䈱䈢䉄䇮䈘䉌䈮Ꮧ ⭯ 䈭䊒䊨䊷䊑ᖱ ႎ 䈎䉌䇮 ㅢ ᖱ ႎ 䈱ⓨ 㑆 ⊛ 䈭 ቢ 䉕ⴕ 䈉ᔅ ⷐ 䈏䈅䉎䇯䈚䈎䈚䇮 䈋䈳䊒䊨䊷 䊑䉦䊷䈱䊥䊮䉪䉦䊋䊷₸ 5䋦䈱 ၞ 䈮ᓥ ᧪ 䈱 䊥䉝䊦䉺䉟 䊛 ቢ ᛛ ⴚ 䉕 ㆡ ↪ 䈜䉎䈫 䇮 ቢ ⚿ ᨐ 䈏ਇ ቯ 䈮䈭䉎䈫䈇䈉 㗴 ὐ 䈏䈅䈦䈢䇯 䈠䈖䈪ᧄ ႎ ๔ 䈪䈲䇮䊒䊨䊷䊑䉦䊷䈱䊥䊮䉪䉦 䊋䊷₸ 䈏ᭂ ┵ 䈮ૐ 䈇႐ ว 䈪䉅 ቯ 䈮䊥䉝䊦䉺 䉟䊛 ቢ 䉕ㆡ ↪ 䈜䉎䈖䈫䉕⋡ ⊛ 䈫䈚䇮 ᴫ 䈱䊒 䊨䊷䊑 ㅢ ᖱ ႎ 䈮ว 䉒䈞䈩േ ⊛ 䈮․ ᓽ ⓨ 㑆 䈱ၮ ᐩ 䉕ㆬ ᛯ 䈜䉎․ ᓽ ⓨ 㑆 ၮ ᐩ ㆬ ᛯ ᚻ ᴺ 䉕 䈮䈧䈇䈩ㅀ 䈼䉎䇯 એ ਅ 䇮2 ┨ 䈪䈲䊔䊷䉴䈫䈭䉎䊥䉝䊦䉺䉟䊛 ቢ ᛛ ⴚ 䈮㑐 䈚䈩䇮ၮ ᧄ ⊛ 䈭䉝䊦䉯䊥䉵䊛䉕⺑ 䈜䉎䇯3 ┨ 䈪䈲䇮䊒䊨䊷䊑䉦䊷䈱䉣䊥䉝䉦䊋 䊷₸ 䈏ૐ 䈇႐ ว 䈻䈱 ᒛ 䉕⋡ ᜰ 䈚䈩䇮േ ⊛ 䈮․ ᓽ ⓨ 㑆 䉕᭴ ᚑ 䈜䉎ᚻ ᴺ 䈮䈧䈇䈩ㅀ 䈼䉎䇯 4 ┨ 䈪䈲䇮䈠䈱ല ᨐ 䉕ᬌ ⸽ 䈜䉎䇯5 ┨ 䈲⚿ ⸒ 䈪䈅䉍䇮 ᓟ 䈱⺖ 㗴 䇮ዷ ᦸ 䈮䈧䈇䈩ㅀ 䈼䉎䇯 −24− 2.
(3) 2. ․ ᓽ ⓨ 㑆 ᓇ 䉕↪ 䈇䈢䊥䉝䊦䉺䉟䊛 ቢ 2.1. 䊥䉝䊦䉺䉟䊛 ቢ 䈱ၮ ᧄ 䉝䊦䉯䊥䉵䊛 䈅䉎න 䉣䊥䉝䈪 㓸 䈘䉏䈢䊒䊨䊷䊑䊂䊷 䉺䉕䇮䊥䊮䉪න 䈱ᣏ ⴕ ᤨ 㑆 䊂䊷䉺䈭䈬䈮ട Ꮏ 䈚䈢 䈪䇮ਥ ᚑ ಽ ಽ ᨆ 䉕ⴕ 䈉䇯䈖䉏䈮䉋䉍䇮 ⶄ ᢙ 䈱䊥䊮䉪䈱䊂䊷䉺䉕䇮⋧ 㑐 䉕䉅䈦䈩ᄌ ൻ 䈜 䉎ᚑ ಽ 䈫䇮ή ⋧ 㑐 䈮ᄌ ൻ 䈜䉎ᚑ ಽ 䈮ಽ ⸃ 䈪䈐 䉎䇯 䈘䉌䈮䇮⋧ 㑐 䈱䈅䉎ᚑ ಽ 䈗䈫䈮䇮න ৻ 䈱ઍ ᄌ ㊂ 䈪 䈜䈖䈫䈏น ⢻ 䈮䈭䉎䈢䉄䇮䊂䊷䉺 䈱ᰴ ᢙ 䈏❗ ㅌ 䈘䉏䉎䇯ᧄ ᧪ 䈱ᣏ ⴕ ᤨ 㑆 䊂䊷 䉺䈲䇮೨ ⸥ ઍ ᄌ ㊂ 䉕ଥ ᢙ 䈫䈚䈩䇮䊥䊮䉪㑆 䈱 ⋧ 㑐 㑐 ଥ 䉕 䈜ၮ Ḱ 䊌䉺䊷䊮䋨䈖䉏䉕ၮ ᐩ 䈫 䈹䋩䉕✢ ᒻ ว ᚑ 䈜䉎䈖䈫䈮䉋䉍䇮ㄭ ૃ ⊛ 䈮 䈘䉏䉎䇯䈖䈱䉋䈉䈮㓸 ⚂ 䈘䉏䈢ᖱ ႎ 䈏䇮․ ᓽ ⓨ 㑆 ᓇ 䈪䈅䉎䇯ၮ ᐩ 䈲․ ᓽ ⓨ 㑆 䉕᭴ ᚑ 䈜䉎㕒 ⊛ 䈭 䊌䊤䊜䊷䉺䈪䈅䉍䇮೨ ⸥ ઍ ᄌ ㊂ 䈏䇮․ ᓽ ⓨ 㑆 䈪േ ⊛ 䈮ᄌ ൻ 䈜䉎ᐳ ᮡ 䈮ኻ ᔕ 䈜䉎䇯 ㅒ 䈮䇮 ᴫ 䈱 ㅢ ᖱ ႎ 䈏䊒䊨䊷䊑䊂䊷䉺 䈱䉋䈉䈮ᄢ 䈐䈭ᰳ ៊ 䉕 䉃䉅䈱䈪䈅䈦䈩䉅䇮䈠 䉏䉕․ ᓽ ⓨ 㑆 䈮 ᓇ 䈜䉎䈖䈫䈏䈪䈐䉏䈳䇮䈠 䈱․ ᓽ ⓨ 㑆 ᐳ ᮡ 䉕ర 䈱 ㅢ ᖱ ႎ 䊂䊷䉺ⓨ 㑆 䈮ㅒ ᓇ 䈜䉎䈖䈫䈪䇮 ㅢ ᖱ ႎ 䈱ᰳ ៊ 䈚䈢䊥 䊮䉪䈮䈧䈇䈩ផ ቯ ቢ 䉕ⴕ 䈉䈖䈫䈏䈪䈐䉎䇯 એ 䉋䉍․ ᓽ ⓨ 㑆 ቢ 䈲䇮Fig. 1䈮␜ 䈜䉋 䈉䈮䇮 䋨䌡䋩ㆊ 䈱䊂䊷䉺䈎䉌․ ᓽ ⓨ 㑆 䉕↢ ᚑ 䈚䇮 䋨䌢䋩䊥䉝䊦䉺䉟䊛䈮ⷰ ᷹ 䈘䉏䈢䊂䊷䉺䈎䉌䇮 ․ ᓽ ⓨ 㑆 䈱ᐳ ᮡ 䉕ቯ 䉄䇮 (c)․ ᓽ ⓨ 㑆 ᐳ ᮡ 䈱ㅒ ᓇ 䈮䉋䈦䈩䇮ផ ቯ ᖱ ႎ 䉕↢ ᚑ 䈜䉎䇮 䈫䈇䈉䋳䈧䈱䊒䊨䉶䉴䈎䉌ᚑ 䉍┙ 䈦䈩䈇䉎䇯એ ਅ 䇮䈠䉏䈡䉏䈱䉴䊁䉾䊒䈮䈧䈇䈩ౕ ⊛ 䈮⺑ 䈜䉎䇯 䉴䊁䉾䊒䋨a䋩 㩷 ․ ᓽ ⓨ 㑆 䈱↢ ᚑ 䈮䈲䇮䇸ᰳ ៊ ୯ ઃ 䈐ਥ ᚑ ಽ ಽ ᨆ 䋨䌐䌃䌁䌍䌄䋩䇹[3][4][5]䉕↪ 䈇䈢䇯䈖䉏䈲 䊒䊨䊷䊑䊂䊷䉺䈲ᄢ ⷙ ᮨ 䈭ᰳ ៊ 䉕 䉃䈢䉄䇮 ㅢ Ᏹ 䈱ਥ ᚑ ಽ ಽ ᨆ 䈲ㆡ ↪ 䈪䈐䈭䈇䈢䉄䈪䈅 䉎䇯 ቢ ኻ ⽎ 䉣䊥䉝䈮䈍䈔䉎 M ᧄ 䈱䊥䊮䉪䈮䈧. 䈇䈩䇮N ࿁ 䈮䉒䈢䈦䈩⸘ ᷹ 䈘䉏䈢 ㅢ ᖱ ႎ 䊂 䊷䉺䉕 N㬍M ⴕ 㪯 䈪 䈜䉅䈱䈫䈜䉎䇯X 䈱 i ⴕ ⋡ 䈱ᚑ ಽ 䉕ኻ ⷺ ⷐ ⚛ 䈫䈜䉎䊂䊷䉺ⴕ D xi 䇮 ㊀ 䉂ⴕ V䇮㩷 V 0 䈮ኻ 䈚䈩䇮䌐䌃䌁䌍䌄䈲䊐䊨 䊔䊆䉡䉴䊉䊦䊛㩷 N. J. ¦ SS Y. i. e M ui. i 1. Yi. T.
(4). Dwi , I. 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩿㪈㪀㩷. D xiV V 0 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩿㪉㪀㩷. 䉕ᦨ ዊ ൻ 䈜䉎 㗴 䈪䈅䉎䇯䈖䈱 㗴 䉕⸃ 䈒䈖䈫 䈪䇮ಣ ℂ ኻ ⽎ 䈱 ㅢ ᖱ ႎ 䊂䊷䉺 X 䈱ⷰ ᷹ ୯ 䉕䇮⺋ Ꮕ 䊉䊦䊛ᦨ ዊ 䈪ㄭ ૃ 䈪䈐䉎ⶄ ᢙ 䈱ၮ ᐩ 䈏ᓧ 䉌䉏䉎䇯䈜䈭䉒䈤䇮 ㅢ ᖱ ႎ 䊂䊷䉺 X 䉕䇮 䌐䌃䌁䌍䌄䈪ᓧ 䉌䉏䈢ၮ ᐩ 䈪ᒛ 䉌䉏䉎․ ᓽ ⓨ 㑆 䈮 ᓇ 䈜䉏䈳䇮䈠䈱ㅒ ᓇ 䈮䉋䈦䈩ਈ 䈋䉌 䉏䉎䊂䊷䉺䈲䇮ర 䈱 ㅢ ᖱ ႎ 䊂䊷䉺䈮ኻ 䈜䉎 ᦨ ዕ ផ ቯ 䈫䈭䉎䇯䈖䈱䈫䈐䇮․ ᓽ ⓨ 㑆 䉕᭴ ᚑ 䈜䉎ၮ ᐩ ᢙ 䉕ᰴ ᢙ 䈫 䈹䇯㩷 㩷 䉴䊁䉾䊒䋨䌢䋩㩷 ࠬ ࠹ ࠶ ࡊ㧔 a㧕ߢ ᓧ ࠄ ࠇ ߚ ၮ ᐩ ߦ ኻ ߒ ߡ ޔ ᰳ៊ߩߥᴫ࠺࠲ࠍᓇߔࠆ႐วߦ ߪޔၮᐩߣᴫ࠺࠲ߩౝⓍߦࠃߞߡޔ ․ ᓽ ⓨ 㑆 ᐳ ᮡ ߪ ৻ ᗧ ߦ ቯ ߐ ࠇ ࠆ ৻ޕᣇ ޔ ᴫ࠺࠲߇ᰳ៊ࠍ߁႐วߦߪޔౝⓍ ߦࠃࠆᓇߪਇน⢻ߢࠅޔ㊀ߺઃߌ ᓇߣ߫ࠇࠆᰴᑼߩ⸃ᴺࠍ↪ࠆޕ a. P W T. T. WP.
(5). 1. P TW TWx T 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩷 㩿㪊㪀. ߎߎߢޔP ߪ㧼㧯㧭㧹㧰ߢᓧࠄࠇߚၮ ᐩࠍਗߴߚⴕߢࠅޔW ߪ㊀ߺઃߌߩ ⴕߢࠆޕᰳ៊ࠍᴫ࠺࠲ x ߦ ኻߒߡޔᓇὐ a ߇ᓧࠄࠇࠆޕ㊀ߺઃߌ ᓇ ߢ ߪ ⷰޔ᷹ ࠺ ࠲ ߩ ㊀ ߺ ࠍ 1ޔᰳ ៊ ࠺ ࠲ߩ㊀ߺࠍ 0 ߣߒߡᛒ߁ߎߣߢޔᰳ៊ ࠺࠲ߩࡦࠢࠍήⷞߒޔᴫ࠺࠲߇ ᷹ⷰߐࠇߚࡦࠢߦߟߡ․ޔᓽⓨ㑆 ߩᓇὐߣޔᓇ೨ߩ࠺࠲ߩ⺋Ꮕࡁ࡞ ࡓ߇ᦨዊൻߐࠇࠆࠃ߁ߦޔᓇὐࠍቯ ߔࠆޔߜࠊߥߔޕ㊀ߺઃߌᓇߦࠃߞߡ ᓧࠄࠇࠆ․ᓽⓨ㑆ᐳᮡߪⷰޔ᷹࠺࠲ߦ ኻߔࠆᦨዕផቯ୯ߢࠆޕ 3 −25−.
(6) 䉴䊁䉾䊒䋨䌣䋩 ࠬ ࠹ ࠶ ࡊ㧔 b㧕ߩ ㊀ ߺ ઃ ߌ ᓇ ߦ ࠃ ߞ ߡ ᓧࠄࠇߚ․ᓽⓨ㑆ᐳᮡ a ࠍޔᰴᑼߦࠃࠅ రߩ࠺࠲ⓨ㑆߳ㅒᓇߔࠆޕ. (4) xˆ aP T ㅒ ᓇ ߢ ᓧ ࠄ ࠇ ߚ xˆ ߪ ․ ޔᓽ ⓨ 㑆 ߩ ᓇὐ߇ x ߦኻߔࠆ⺋Ꮕࡁ࡞ࡓᦨዊ⸃ߢ ࠆ ߣ ߁ ᕈ ⾰ ߆ ࠄ ޔx ߩ ⷰ ᷹ ୯ ߦ ኻ ߒ ߡ ߪ ߘߩㄭૃ୯ߢࠆ․ߚ߹ޕᓽⓨ㑆߇ࡦ ࠢ 㑆 ߩ ⋧ 㑐 㑐 ଥ ࠍ ߔ ߎ ߣ ߆ ࠄ ޔx ߩ ᰳ ៊ ୯ߦኻߔࠆផቯ୯ߢࠆޕx ߩᰳ៊୯ࠍ xˆ ߢ ⟎ ߈ ឵ ߃ ࠆ ߎ ߣ ߢ ޔx ߩ ቢ ߇ ὑ ߐ ࠇ ࠆޕ. 2.2. Ꮧ ⭯ ⁁ ᴫ ਅ 䈮䈍䈔䉎 㗴 ὐ ࡊࡠࡉࠞ߇Ꮧ⭯ߦሽߔࠆ⁁ᴫߦ ߅ߡޔ೨▵ߢ⸃⺑ߒߚ․ᓽⓨ㑆ቢߩ 㗴ὐߦߟߡㅀߴࠆޕᴫߩࡊࡠࡉ ࠺࠲ࠍ㓸ߢ߈ߚࡦࠢ㧔એਅࡠࡊޔ ࡉ࠺࠲᷹ⷰࡦࠢ㧕ᢙ߇ᭂ┵ߦዋߥ ႐วޔℂ⺰⊛ߦផቯቢߩ⚿ᨐࠍ᳞ ࠆߎߣ߇ߢ߈ߥߪߚ߹ޔផቯቢߩ⚿ ᨐ߇ਇቯߦߥࠆߣ߁㗴ὐ߇↢ߓࠆޕ એਅⷰ࠲࠺ࡉࡠࡊޔ᷹ࡦࠢᢙߣ․ ᓽⓨ㑆ߩᰴᢙߣߩᄢዊ㑐ଥࠍ႐วಽߌߒޔ ࡊࡠࡉ࠺࠲᷹ⷰࡦࠢᢙ߇ᭂ┵ߦዋ ߥ႐วߩ㗴ὐߦߟߡ⺑ߔࠆޕ ࡊࡠࡉ࠺࠲᷹ⷰࡦࠢᢙࠃࠅ߽․ᓽ ⓨ㑆ᰴᢙ߇ᄙ႐ว ℂ⺰⊛ߦផቯቢߩ⚿ᨐࠍ᳞ࠆߎߣ ߇ ߢ ߈ ߥ ߪ ࠇ ߎޕᑼ ߩ ⴕ 㧼 T 㨃 T 㨃 㧼߇❗ㅌߒߩߘޔㅒⴕࠍዉߢ߈ߥ ߚߢࠆޕ ࡊࡠࡉ࠺࠲᷹ⷰࡦࠢᢙࠃࠅ߽․ᓽ ⓨ㑆ᰴᢙ߇ዋߥ႐ว ℂ⺰⊛ߦផቯቢߩ⚿ᨐࠍዉߢ߈ࠆ ߇⇣ޔᏱߥᴫ࠺࠲߇ജߐࠇࠆߣޔ ߘߩ࠺࠲ߦోࡦࠢߩജ⚿ᨐ߇ᄢ߈ ߊᓇ㗀ߒޔផቯቢߩ♖ᐲ߇ᖡߊߥࠆน ⢻ᕈ߇ࠆޕ㜞♖ᐲߥផቯቢࠍታߔ. feature space. link3 2nd㩷 basis. 1st basis link2 link1 Fig. 2. Example of feature space. ࠆߚߦߪ․ޔᓽⓨ㑆ᰴᢙࠃࠅ߽ࡠࡊޔ ࡉ࠺࠲᷹ⷰࡦࠢᢙ߇ࠆ⒟ᐲએޔ ᄙᔅⷐ߇ࠆޕ ೨ㅀߒߚ․ᓽⓨ㑆ቢߢߪ․ޔᓽⓨ㑆 ߩ᭴ᚑࠍࠝࡈࠗࡦߢⴕߞߡߚߩߎޕ ߚ․ޔᓽⓨ㑆ߩᰴᢙߪࠝࡈࠗࡦ⸘▚ ᤨ ߦ ቯ ߐ ࠇ ࠆ ౕ ޕ ⊛ ߦ ߪ ޔᑼ (1) ߆ ࠄ ዉߒߚⶄᢙߩၮᐩߩነਈ₸ߩ㜞㗅ߦ ၮᐩࠍߔࠆߩߎޕฦၮᐩߩነਈ₸ߪ ၮᐩߩᖱႎ㊂ࠍߒߡ߅ࠅޔߩၮᐩ ߪቢኻ⽎ߣߔࠆࡦࠢߦ߅ߡਥⷐߥ ㅢᖱႎߩᄌൻࠍߒޔਅߩၮᐩߪシ ᓸߥㅢᖱႎߩᄌൻࠍߒߡࠆޕ ߎߩߚޔシᓸߥㅢᖱႎߩᄌൻ߹ߢ ࠍࠕ࡞࠲ࠗࡓߢቢߒߚ႐วߪࠝޔ ࡈࠗࡦ⸘▚ᤨߦ㜞ᰴᢙߩ․ᓽⓨ㑆ࠍ ᭴ᚑߔࠆᔅⷐ߇ࠆޔߒ߆ߒޕᴫߩࡊ ࡠࡉ࠺࠲᷹ⷰࡦࠢᢙ߇㕖Ᏹߦዋߥ ႐วߦߪޔផቯቢ⚿ᨐ߇ਇቯ߹ߚ ߪዉߢ߈ߥน⢻ᕈ߇ࠆޕ ߎߩ㗴ࠍ࿁ㆱߔࠆߚߦࠗࡈࠝޔ ࡦ⸘▚ᤨߦૐᰴᢙߩ․ᓽⓨ㑆ࠍ᭴ᚑߒ ߚ႐วޔᴫߩࡊࡠࡉ࠺࠲᷹ⷰࡦ ࠢᢙ߇ᄙ႐วߢߞߡ߽ޔਥⷐߥㅢ ᖱႎߩᄌൻߛߌߒ߆ቢߢ߈ߕޔචಽߦ ᴫߩࡊࡠࡉ࠺࠲ࠍᵴ߆ߔߎߣ߇ߢ ߈ߥⷰ࠲࠺ࡉࡠࡊޔߦࠄߐޕ᷹ ࡦࠢᢙ߇ዋߥ႐วߦ߽ߦࠢࡦߩߘޔ ⋧㑐ߩࠆᖱႎ߇ޔዋߥᰴᢙߩ․ᓽⓨ 㑆ߢߐࠇࠆߣߪ㒢ࠄߥޕ. −26− 4.
(7) 3. ․ ᓽ ⓨ 㑆 䈱േ ⊛ ၮ ᐩ ㆬ ᛯ ᚻ ᴺ. link2. 3.1. ၮ ᧄ ᔨ ᴫ࠺࠲ߩ⋧㑐ߩᒝၮᐩࠍࠕ࡞ ࠲ࠗࡓߦㆬᛯߒ․ޔᓽⓨ㑆ࠍേ⊛ߦ᭴ᚑ ߔࠆߎߣߦࠃࠅࠕࠛߩࠞࡉࡠࡊޔ ࠞࡃ₸ߦᔕߓߡߩ࠲࠺ࡉࡠࡊޔⓨ 㑆⊛ߥᰳ៊ࠍቢߔࠆᚻᴺࠍㅀߴࠆޕ ․ᓽⓨ㑆ࠍേ⊛ߦ᭴ᚑߔࠆߣߪޔ೨▵ 2.1 ߩ ࡊ ࡠ ࠬ (a) ߢ ▚ ߐ ࠇ ࠆ ⶄ ᢙ ߩ ၮ ᐩࠍޔᴫߩࡊࡠࡉ࠺࠲ߦวࠊߖߡ ㆬ ᛯ ߒ ․ޔᓽ ⓨ 㑆 ࠍ ᭴ ᚑ ߔ ࠆ ߎ ߣ ߢ ࠆ ޕ ၮᐩߪⷰ࠲࠺ࡉࡠࡊޔ᷹ࡦࠢߩ ⋧㑐ࠍᒝߊߒߡࠆ߽ߩࠍㆬᛯߔࠆޕ ․ ᓽ ⓨ 㑆 ߩ ৻ ࠍ Fig.2 ߦ ␜ ߔ ޕၮ ᐩ ߪ ฦ ࡦࠢߦ߅ߡ⋧㑐ࠍᜬߞߡᄌൻߔࠆ ㅢᖱႎߩᚑಽߢ᭴ᚑߐࠇࠆ߫߃ޕၮᐩ 㧝ߦ߅ߌࠆࡦࠢ㧝ࠢࡦޔ㧞ࠢࡦޔ 㧟 ߩ ߘ ࠇ ߙ ࠇ ߩ ᚑ ಽ ࠍ [ l 11 , l 12 , l 13 ]=[ 0.1, 1.0, 0.2 ] ߣ ߔ ࠆ ߣ ࠢ ࡦ ޔ㧝 㨪 㧟 ߩ ㅢ ᖱ ႎ ߦ ”㧝 㧦 㧝 㧜 㧦 㧞 ”ߣ ߁ Ყ 㑐 ଥ ߢ ᄌൻߔࠆᚑಽ߇߹ࠇߡࠆߎߣࠍᗧ ߒ ߡ ࠆ ޕၮ ᐩ 㧞 ߩ ᚑ ಽ ߪ [ l 21 , l 22 , l 23 ]=[ 0.1, 0.2, 1.0 ] ߣ ߔ ࠆ ޕFig.2 ߩ ၮ ᐩ 㧝 ߪࠢࡦޔ㧞ߩ⋧㑐ࠍᒝߊߒߡࠆޕ ߎߩߚ߃߫ޔၮᐩࠍ㧝ᧄㆬᛯߔࠆ⁁ ᴫߢࠢࡦޔ㧞ߩߺᴫࡊࡠࡉ࠺࠲ ࠍ㓸ߢ߈ߚ႐วߪޔၮᐩ㧝ࠍㆬᛯߒޔ ․ᓽⓨ㑆ࠍ᭴ᚑߔࠆᧄޕႎ๔ߢߪޔᴫ ߩࡊࡠࡉ࠺࠲ߩ⋧㑐ߩᒝߐࠍޔᴫ ࠺࠲ࠍฦၮᐩ߳ߩᓇߒߚᓇࡌࠢ࠻ ࡞ߩࡁ࡞ࡓߢ⹏ଔߔࠆޕ. 3.2. ․ ᓽ ⓨ 㑆 䈱േ ⊛ ၮ ᐩ ㆬ ᛯ 䈱䊒䊨䉶䉴 േ⊛ߥၮᐩㆬᛯᚻᴺߪޔ (i) ᴫ ࡊ ࡠ ࡉ ᖱ ႎ ࠍ ฦ ၮ ᐩ ߦ ᓇ ޔ (ii) ᓇ ߒ ߚ ᓇ ࡌ ࠢ ࠻ ࡞ ߩ ࡁ ࡞ ࡓ ࠍ ޔฦ ၮᐩߩಽᢔߢ㊀ߺઃߌࠍߒޔฦၮᐩߩ⹏ ଔ୯ࠍ▚ޔ (iii) ߘ ߩ ⹏ ଔ ୯ ߩ 㜞 㗅 ߦ ࠺ ࡉ ࡠ ࡊޔ ࠲᷹ⷰࡦࠢᢙߦᔕߓߚᢙߩၮᐩࠍㆬᛯޔ (iv) ㆬ ᛯ ߐ ࠇ ߚ ၮ ᐩ ߆ ࠄ ․ ᓽ ⓨ 㑆 ࠍ ᭴ ᚑ ߒޔᴫ࠺࠲ࠍㅒᓇߒߡޔផቯቢ. i-th pi si. d Fig. 3. projection. >1 0@T. link1. Evaluation Index.. ᖱႎࠍ↢ᚑߔࠆޔ ߣ߁ࡊࡠࠬ߆ࠄᚑࠅ┙ߟޕ ߎߩၮᐩㆬᛯᚻᴺߪߩࠞࡉࡠࡊޔ ࡦࠢࠞࡃ₸ߦᔕߓߡޔၮᐩߩ㓸ว߆ ࠄ ㆬ ᛯ ߔ ࠆ ၮ ᐩ ᢙ ࠍ น ᄌ ߦ ߔ ࠆ ޔ ߡ ߞ ࠃޕ ࡊࡠࡉࠞ߇Ꮧ⭯ߦሽߔࠆ⁁ᴫߛߌ ߢߥߊࠢࡦߩࠞࡉࡠࡊࠆࠁࠄޔ ࠞࡃ₸ߦߟߡㆡ↪ߢ߈ࠆޕ ࠞࡃ₸߇㜞႐วߪߢࡦࠗࡈࠝޔ ᚑߒߚၮᐩߩ㓸ว߆ࠄޔᄙᢙߩၮᐩࠍ ㆬᛯߒߡ․ᓽⓨ㑆ࠍ᭴ᚑߒޔផቯቢࠍ ⴕ߁ޕᄙߊߩၮᐩࠍ↪ࠆߎߣߢࡦޔ ࠢ㑆ߩシᓸߥㅢᖱႎߩᄌൻ߹ߢቢߢ ߈♖ޔᐲߩ㜞ផቯቢࠍታߢ߈ࠆޕ ৻ᣇ߇₸ࡃࠞࠢࡦޔૐ႐วߪޔዋ ᢙߩၮᐩࠍㆬᛯߒߡ․ᓽⓨ㑆ࠍ᭴ᚑߔࠆ ߎߣߢࠍ࠲࠺ޔ㓸ߒߚࡦࠢߩ⋧㑐 ߩᒝ․ᓽⓨ㑆ߢផቯቢࠍታߢ߈ࠆޕ. 3.3. ․ ᓽ ⓨ 㑆 䈱േ ⊛ ၮ ᐩ ㆬ ᛯ 䈱 ⚦ ᴫ 䊒䊨䊷 䊑䊂䊷䉺䈱䊥䊮䉪ᖱ ႎ 䉕 䇮ฦ ၮ ᐩ 䈻䈱 ᓇ 䈜䉎䋨೨ ▵ 䋳䋮㪉 䈱䊒䊨䉶䉴㩿㫀㪀䋩䇯 ቢ ኻ ⽎ 䈱䉣䊥䉝䈮䈍䈔䉎 M ᧄ 䈱䊥䊮䉪䈮䈍䈇 䈩 ᴫ 䊒䊨䊷䊑䊂䊷䉺䈱䊥䊮䉪ᖱ ႎ 䊔䉪䊃䊦 d 䉕㩷. d. >d1. d2 dM @ 㩷 T. 㩿㪌㪀. 䈫䈜䉎䇯䈖䈱 d i 䈲䇮i ⇟ ⋡ 䊥䊮䉪䈮䈍䈇䈩䊒䊨䊷 䊑 ㅢ ᖱ ႎ 䉕 㓸 䈪䈐䈢႐ ว 䈲 1 䇮 㓸 䈪䈐 䈝ᰳ ៊ 䈚䈩䈇䉎႐ ว 䈲 0 䈱୯ 䉕䈫䉎䇯 䈋䈳䇮. 5 −27−.
(8) Fig. 5 Example of imputation. Fig. 4. Evaluation area.. 䊒䊨䊷䊑䉦䊷䈎䉌䊥䊮䉪 㪈䇮䊥䊮䉪 㪉 䈱䊒䊨䊷䊑 ㅢ ᖱ ႎ 䉕 㓸 䈪䈐䇮䊥䊮䉪 㪊 䈱 ㅢ ᖱ ႎ 䈏 㓸 䈪䈐䈝ᰳ ៊ 䈚䈩䈇䉎䈫䈐䇮 ᴫ 䊒䊨䊷䊑 䊂䊷䉺䈱䊥䊮䉪ᖱ ႎ 䊔䉪䊃䊦䈲 d 䋽 [ 1 1 0 ] T 䈫䈭䉎䇯㩷 ᴫ 䊒䊨䊷䊑䊂䊷䉺䈱䊥䊮䉪ᖱ ႎ 䊔䉪䊃䊦 d 䉕 i ⇟ ⋡ 䈱ၮ ᐩ 䊔䉪䊃䊦 p i 䈻 ᓇ 䈚䈢䈫䈐䇮 ၮ ᐩ 䊔䉪䊃䊦ⓨ 㑆 䈮䈍䈔䉎 ᓇ ὐ ᐳ ᮡ t i 䈲䇮 ᴫ 䊂䊷䉺䊔䉪䊃䊦 d 䈫䈱ౝ Ⓧ 䈎䉌䇮㩷. t. T. i. pi d. 㩿㪍㪀. 䈪䈅䉎䇯䈖䉏䉕ర 䈱䊥䊮䉪ᐳ ᮡ ♽ 䈪 䈜䈫䇮㩷. si. p i p iT d 㩷. 㩿㪎㪀. 䈫䈭䉍䇮i ⇟ ⋡ 䈱ၮ ᐩ 䊔䉪䊃䊦 p i 䈻䈱 ᓇ 䊔䉪 䊃䊦䈮䈭䉎䇯 Fig. 3 䈲䇮䊥䊮䉪䋱䈫䊥䊮䉪䋲䈱䊥䊮䉪 ᐳ ᮡ ♽ 䈱 ᴫ 䊂䊷䉺䊔䉪䊃䊦 d 䋽 [ 1 0 ] T 䉕 i ⇟ ⋡ 䈱ၮ ᐩ 䊔䉪䊃䊦 p i 㩷 䈻 ᓇ 䈘䈞䈢 䈪 䈅䉎䇯㩷 ᰴ 䈮䇮ฦ ၮ ᐩ 䈱 ᓇ 䊔䉪䊃䊦䉕↪ 䈇䈩䇮ฦ ၮ ᐩ 䈱⹏ ଔ ୯ 䉕▚ 䈜䉎䋨೨ ▵ 䋳䋮㪉 䈱䊒䊨䉶 䉴㩿㫀㫀㪀䋩䇯೨ ㅀ 䈪᳞ 䉁䈦䈢 ᓇ 䊔䉪䊃䊦 s i 䈱䊉䊦 䊛䈲䇮i ⇟ ⋡ 䈱ၮ ᐩ 䈫䇮 ᴫ 䊒䊨䊷䊑䊂䊷䉺䉕 㓸 䈚䈢䊥䊮䉪⟲ 䈫䈱⋧ 㑐 䈱ᒝ 䈘䉕 䈚䈩䈇䉎䇯 䈖䈱 ᓇ 䊔 䉪䊃䊦䈱䊉䊦䊛䉕↪ 䈇䈩ฦ ၮ ᐩ 䈱 ⹏ ଔ ୯ 䉕▚ 䈜䉎䇯i ⇟ ⋡ 䈱ၮ ᐩ 䊔䉪䊃䊦 p i 䈱⹏ ଔ ୯ v i 䈲㩷. vi. Oi s i. Oi p i p iT d 㩷. 㩿㪏㪀. 䈫䈜䉎䇯㱗 i 䈲 㪧㪚㪘㪤㪛 䈱ㆊ ⒟ 䈪䇮ၮ ᐩ 䈫ኻ 䈪 ᓧ 䉌䉏䉎࿕ ୯ 䈪䈅䉍䇮․ ᓽ ⓨ 㑆 䈱╙ i ゲ 䈮 ᴪ 䈦䈢䊂䊷䉺䈱ಽ ᢔ 䉕 䈚䈩䈇䉎䇯㱗 i 䉕ᱜ ⷙ ൻ 䈚䈢୯ 䈏ၮ ᐩ i 䈱ነ ਈ ₸ 䈪䈅䉎䈖䈫䈎䉌䇮ᑼ 㩷 㩿㪏㪀䈲䇮 ᓇ 䊔䉪䊃䊦䈱䊉䊦䊛䉕ነ ਈ ₸ 䈪㊀ 䉂 ઃ 䈔䈚䈢୯ 䈪䈅䉎䇯䈖䈱⹏ ଔ ୯ 䉕↪ 䈇䉎䈖䈫䈪䇮 ฦ ၮ ᐩ 䈫 䇮 ᴫ 䊒 䊨 䊷 䊑 䊂 䊷 䉺 䉕 㓸 䈚 䈢䊥 䊮䉪⟲ 䈫䈱⋧ 㑐 䈱ᒝ 䈘䉕⹏ ଔ 䈜䉎䈖䈫䈏䈪䈐䉎䇯 䈖䈱⹏ ଔ ୯ 䈱㜞 䈇㗅 䈮 䌎 䌐 䈱ၮ ᐩ 䉕ㆬ ᛯ 䈚 䋨೨ ▵ 䋳䋮㪉 䈱䊒䊨䉶䉴㩿㫀㫀㫀㪀䋩䇮ᣂ 䈚䈇․ ᓽ ⓨ 㑆 䉕᭴ ᚑ 䈜䉎䋨೨ ▵ 䋳䋮㪉 䈱䊒䊨䉶䉴㩿㫀㫍㪀䋩䇯 䌎 䌐 䈲 䊒䊨䊷䊑䉦䊷䈱䊥䊮䉪䉦䊋䊷₸ 䈮䉋䉍 ቯ 䈘䉏 䉎䇯㩷 㩷 4. ․ ᓽ ⓨ 㑆 䈱േ ⊛ ၮ ᐩ ㆬ ᛯ ᚻ ᴺ 䈱ᬌ ⸽ 4.1. ᬌ ⸽ ᚻ 㗅 ᧄႎ๔ߢߪޔታ㓙ߩࡊࡠࡉ࠺࠲߆ ࠄࡦࠢᣏⴕᤨ㑆ࠍੱὑ⊛ߦᰳ៊ߐߖޔ ߘߩ࠺࠲ࠍ↪ߡឭ᩺ᚻᴺߩᬌ⸛ࠍⴕ ߁ޕ 㧔 1㧕 ᬌ ⸽ ߦ ߪ ᧲ ੩ ㇺ ౝ 2 ᰴ ࡔ ࠶ ࠪ ࡘ 533935 ౝ ߦ ߅ ߌ ࠆ ࠲ ࠢ ࠪ ߩ ࡊ ࡠ ࡉ ࠺ ࠲ ࠍ ↪ ߚ ޕFig. 4 ߦ ⹏ ଔ ࠛ ࠕ ࠍ ␜ ߔ ޕ ࠺ ࠲ ߩ ⫾ Ⓧ ᦼ 㑆 ߪ 1 ࡩ ಽ㧔 2005 ᐕ 10 1 ᣣ 㨪 31 ᣣ 㧕 ߢ ࠆ ࡠ ࡊ ߩ ࠪ ࠢ ࠲ ޕ ࡉ ࠺ ࠲ ࠍ ࿑ ߦ ࡑ ࠶ ࠴ ࡦ ࠣ ߒ ޔ5 ಽ ߅ ߈ߩࡦࠢᣏⴕᤨ㑆ߦᄌ឵ߒߚߢߎߎޕ. 6 −28−.
(9) 4.2. ᬌ ⸽ ⚿ ᨐ Fig. 5 ߪ 2 ᰴ ࡔ ࠶ ࠪ ࡘ 533935 ౝ ߩ 㧝 ᧄ ߩࡦࠢߩផቯቢ⚿ᨐߢࠅ࠳ࡦޔ ࡓᰳ៊ߐߖߚᴫ࠺࠲ޔర࠺࠲ޔឭ ᩺ߒߚᚻᴺߩቢ⚿ᨐޔᓥ᧪ߩ⛔⸘୯ߦ ࠃࠆቢ⚿ᨐࠍᤨ♽㗅ߦ␜ߒߡࠆޕ ❑ ゲ ߪ ࡦ ࠢ ᣏ ⴕ ᤨ 㑆 ޔᮮ ゲ ߪ 2005 ᐕ 10 18 ᣣ ඦ ೨ 6 ᤨ ߆ ࠄ 19 ᣣ ඦ ᓟ 0 ᤨ ߹ ߢ ᤨ♽࠺࠲ࠍߒߡࠆޕ 2005 ᐕ 10 15 ᣣ ߆ ࠄ 31 ᣣ ߦ ߅ ߡ ޔ ࡦࠢᲤߢቢ⺋Ꮕߩᐔᣇੑਸ਼ᐔဋ. 㪽㫉㪼㫈㫌㪼㫅㪺㫐. ⹏ ଔ ኻ ⽎ ߩ ࡦ ࠢ ߪ ޔ2 ᰴ ࡔ ࠶ ࠪ ࡘ 533935 ౝ ߩ ਥ ⷐ ߥ ࡦ ࠢ 598 ᧄ ߢ ࠆ ޕ 㧔 2 㧕 2005 ᐕ 10 1 ᣣ ߆ ࠄ 2 ㅳ 㑆 ಽ ߩ ࠺࠲ࠍ↪ߡ․ᓽⓨ㑆ߩၮᐩࠍ PCAMD ࠍ ↪ ߡ ▚ ߔ ࠆ ޔ ߚ ߹ޕᲧ セ ߩ ߚ ߦ ⸘ ⛔ޔ୯ ࠍ ↪ ߚ ᓥ ᧪ ᚻ ᴺ ߣ ߒ ߡ ޔ หᦼ㑆ߦ߅ߡหᤨೞᐔဋࠍ⸘▚ߔࠆޕ 㧔 3㧕 ᚻ 㗅 㧔 2㧕 ߢ ᓧ ࠄ ࠇ ߚ ၮ ᐩ ࠍ ↪ ߡޔᏗ⭯⁁ᴫߢߩផቯቢࠍታߔࠆޕ ផቯቢኻ⽎ߣߥࠆࡊࡠࡉ࠺࠲ߪޔ 2005 ᐕ 10 15 ᣣ ߆ ࠄ 31 ᣣ ߩ ࡦ ࠢ ᣏ ⴕ ᤨ㑆࠺࠲ߢࠆࠞࡉࡠࡊߢߎߎޕ ߩ Ꮧ ⭯ ߥ ⁁ ᴫ ࠍ ࠅ ߔ ߚ ߦ ޔ5 ಽ න ߩ ࡦ ࠢ ࠞ ࡃ ₸ ߇ 5 㧑 ߦ ߥ ࠆ ࠃ ߁ ߦ ޔ ࡦࠢᣏⴕᤨ㑆࠺࠲ࠍࡦ࠳ࡓߦᰳ៊ߐ ߖߚޕ 㧔 4㧕 ᚻ 㗅 㧔 3㧕 ߢ ᚑ ߒ ߚ Ꮧ ⭯ ߥ ࡦ ࠢ ᣏⴕᤨ㑆࠺࠲ࠍ 5 ಽᲤߦផቯቢߔࠆޕ ജߐࠇߚᏗ⭯ߥࡦࠢᣏⴕᤨ㑆࠺࠲ ߆ࠄേ⊛ߦၮᐩࠍㆬᛯߒ․ޔᓽⓨ㑆ࠍ᭴ ᚑߔࠆ߈ߣߩߎޕㆬᛯߔࠆၮᐩᢙߪ 5 ߣ ߒ ߚ㧔 N P =5 㧕ޕ᭴ ᚑ ߒ ߚ ․ ᓽ ⓨ 㑆 ߦ ᴫ ࠺ ࠲ ࠍ ᓇ ߒ ޔផ ቯ ቢ ⚿ ᨐ ࠍ ▚ ߔ ࠆ ޕ 㧔 5㧕 ᚻ 㗅 㧔 4㧕 ߢ ജ ߒ ߚ ቢ ࠺ ࠲ ߩ⺋Ꮕ⹏ଔࠍⴕ߁߈ߣߩߎޕᲧセኻ⽎ߪ ࡦ ࠳ ࡓ ᰳ ៊ ೨ ߩ 2005 ᐕ 10 15 ᣣ ߆ ࠄ 31 ᣣ ߹ ߢ ߩ ࡦ ࠢ ᣏ ⴕ ᤨ 㑆 ࠺ ࠲ ߢ ࠆ ޕ ৻ᣇߢߪ⸘⛔ޔ୯ߦࠃࠆቢߩ⺋Ꮕ⹏ଔ ߽ⴕᧄޔႎ๔ߩᛛⴚߩផቯቢ♖ᐲߣ ߩᲧセࠍⴕ߁ޕ. 㪉㪇㪇 㪈㪏㪇 㪈㪍㪇 㪈㪋㪇 㪈㪉㪇 㪈㪇㪇 㪏㪇 㪍㪇 㪋㪇 㪉㪇 㪇. 㫇㫉㫆㫇㫆㫊㪼㪻 㫋㫉㪸㪻㫀㫋㫀㫆㫅㪸㫃. 㪇 㪇㪅㪉 㪇㪅㪋 㪇㪅㪍 㪇㪅㪏 㪈 㪈㪅㪉 㪈㪅㪋 㪈㪅㪍 㪈㪅㪏 㪉. 㪩㪤㪪㪜. Fig. 6 Evaluation error. 㧔 RMSE 㧕ࠍ ▚ ߒ ߚ ޕേ ⊛ ߦ ․ ᓽ ⓨ 㑆 ࠍ ᭴ ᚑ ߔ ࠆ ࠕ ࡞ ࠲ ࠗ ࡓ ቢ ⚿ ᨐ ߩ RMSE ߪ 0.45 ޔᓥ ᧪ ߩ ห ᤨ ೞ ᐔ ဋ ߩ ⛔ ⸘ ୯ ߦ ࠃ ࠆ ቢ ⚿ ᨐ ߩ RMSE ߪ 0.62 ߣ ߥ ߞ ߚ ߐޕ ࠄߦၮᐩㆬᛯߦࠃࠆࠕ࡞࠲ࠗࡓቢ⚿ ᨐߣ⛔⸘୯ߩቢ⚿ᨐࠍᲧセߒߚࡦࠢ ߩ RMSE ߩ ࡅ ࠬ ࠻ ࠣ ࡓ ࠍ Fig.6 ߦ ␜ ߔ ޕ Fig.6 ߦ ␜ ߔ ㅢ ࠅ ⸘ ⛔ޔ୯ ߦ ࠃ ࠆ ቢ ⚿ ᨐ ߪ RMSE ߇ 1.0 ࠍ ߃ ߚ ▸ ࿐ ߦ ߽ ᐢ ߊ ಽ Ꮣߒߡࠆߩߦኻߒޔၮᐩㆬᛯߦࠃࠆ ࠕ ࡞ ࠲ ࠗ ࡓ ቢ ⚿ ᨐ ߪ ޔRMSE ߇ 0.3 ߆ ࠄ 0.9 ߩ ▸ ࿐ ߦ ಽ Ꮣ ߒ ߡ ࠆ ߎ ߣ ߇ ࠊ ߆ ࠆ ޕ એ ࠃ ࠅ ⛘ ޔኻ ⊛ ߥ RMSE ߩ ୯ ߇ ᄢ ߈ ߽ ߩ ߩ ࠢ ࡦ ޔᲤ ߩ RMSE ߪ ޔᓥ ᧪ ߩ ቢ ᚻ ᴺ 0.62 ߆ ࠄ ၮ ᐩ ㆬ ᛯ ߩ ࠕ ࡞ ࠲ ࠗ ࡓ ቢ ᚻ ᴺ 0.45 ߳ ߣ ♖ ᐲ ߇ ะ ߒ ߡ ߅ ࠅ ޔ ৻ቯߩലᨐ߇ᓧࠄࠇߚߣ⸒߃ࠆޕ RMSE ߩ ୯ ߇ ᄢ ߈ ℂ ↱ ߪ ࡉ ࡠ ࡊ ޔ ࠺࠲߆ࠄᚑߒߚࡦࠢᣏⴕᤨ㑆࠺ ࠲ߩ߫ࠄߟ߈ߩᓇ㗀ࠍฃߌߡࠆߚߣ ⠨ ߃ ࠄ ࠇ ࠆ ޕFig. 5 ߩ ޟᰳ ៊ ೨ ߩ ࡊ ࡠ ࡉ ࠺࠲⌀ޔߦ߁ࠃࠆࠇࠄߢޠ୯ߢࠆ ࡊࡠࡉ࠺࠲ߪ߫ࠄߟߡሽߔࠆߚ ޔRMSE ߽ ᄢ ߈ ߥ ୯ ߦ ߥ ࠆ ࡦ ߦ ․ ޕ ࠢᣏⴕᤨ㑆ߩ߫ࠄߟ߈ߪޔరߦߥࠆࡊࡠ ࡉࠞߩบᢙ߇ዋߥ႐วޔዋߥࡊ ࡠࡉࠞߩേߦࠢࡦޔᣏⴕᤨ㑆߇. −29− 7.
(10) ᓇ 㗀 ߐ ࠇ ޔା ภ ߩ ᓇ 㗀 ╬ ߇ 㗼 ⪺ ߦ ࠇ ࠆ ޕ ᓟࡊࡠࡉ࠺࠲ߩ߫ࠄߟ߈ࠍ⠨ᘦߒ ߚ⺋Ꮕ⹏ଔᣇᴺߩᬌ⸛ߩᔅⷐ߇ࠆޕ. 5. ⚿ ⸒ ᧄ ⎇ ⓥ 䈪䈲䇮䊒䊨䊷䊑䉦䊷䈏Ꮧ ⭯ 䈮ሽ 䈜䉎⁁ ᴫ ਅ 䈮䈩䇮䊒䊨䊷䊑䉦䊷䈱ⓨ 㑆 ⊛ 䈭ᰳ ៊ 䉕 ቢ 䈜 䉎䈖䈫䉕⋡ ⊛ 䈫䈚䈩䇮․ ᓽ ⓨ 㑆 䈱 േ ⊛ ᭴ ᚑ 䈮䉋䉎䊥䉝䊦䉺䉟䊛 ቢ ᛛ ⴚ 䉕ឭ ᩺ 䈚䈢䇯䉁䈢ታ 㓙 䈱䉺䉪䉲䊷䈱䊒䊨䊷䊑䊂䊷䉺䉕 ↪ 䈇䈩䇮ឭ ᩺ 䈚䈢ᛛ ⴚ 䈱ᬌ ⸽ 䉕ⴕ 䈇䇮ᓥ ᧪ 䈱 ⛔ ⸘ ୯ 䉕 ↪ 䈇䈢 ቢ ᛛ ⴚ 䉋 䉍䉅 ቢ ♖ ᐲ 䈏 ะ 䈜䉎䈖䈫䉕⏕ 䈚䈢䇯 䈭䈍䇮ᧄ ႎ ๔ 䈪䈲ㆊ 䈱䊒䊨䊷䊑ᖱ ႎ 䉕 ච ಽ 䈮⫾ Ⓧ 䈜䉎䈖䈫䈪䇮䊥䊮䉪㑆 䈱 ⋧ 㑐 㑐 ଥ 䉕 䈚䈩䈇䉎ၮ ᐩ 䉕䇮ච ಽ 䈮▚ 䈚䈩䈇䉎䈖䈫 䉕೨ ឭ 䈮䈚䈩䈇䉎䇯 ᓟ 䈲䇮䊒䊨䊷䊑䉦䊷䈱 ㆊ 䊂䊷䉺䈎䉌 PCAMD 䉕↪ 䈇䈩ၮ ᐩ 䉕▚ 䈜䉎㓙 䈮䇮䈬䈱䈒䉌䈇䈱ᦼ 㑆 䈱ㆊ 䊂䊷䉺 䈏䈅䉏䈳䇮䊒䊨䊷䊑䉦䊷䈏Ꮧ ⭯ 䈮ሽ 䈜䉎⁁ ᴫ 䈪䉅䇮ච ಽ 䈭ၮ ᐩ 䉕▚ 䈜䉎䈖䈫䈏䈪䈐䉎 䈱䈎ᬌ ⸽ 䈜䉎੍ ቯ 䈪䈅䉎䇯 ৻ ᣇ 䇮 䇮ᣣ ┙ ᚲ 䈲䇮⚻ ᷣ ↥ ᬺ ⋭ 䈱ᜰ ዉ 䈪䇮᧲ ੩ 23 䉕ኻ ⽎ 䈮䈚䈢䊒䊨䊷䊑 ㅢ ᖱ ႎ 䊒䊤䉾䊃䊐䉤䊷䊛䈱㐿 ⊒ 䊒䊨䉳䉢䉪䊃 䋨 COSE 䋩䈮䉅ෳ ട 䈚䈩 䈍䉍䇮䈠䈱ታ ↪ ൻ 䈮ᒰ 䈢䈦䈩䉅䇮ᧄ ⎇ ⓥ 䈱ᚑ ᨐ 䈲ነ ਈ 䈪䈐䉎䈫⠨ 䈋 䉎䇯. of Real-Time Floating Car Data Based on Multiple Link Correlation in Feature Space”, Proc. of 13th World Congress on ITS London, CD-ROM, Oct. 2006. [3] A. Ruhe, “Numerical computation of principal components when several observations are missing”, Tech Rep. UMINF-48,Dept.Information Processing, Umea Univ., 1974. [4] ᩊ ጊ “ ޔᰳ ៊ ୯ ߇ ࠆ ႐ ว ߩ ✢ ᒻ ╬ ൻ ᴺ ”ޔᢎ ⢒ ᔃ ℂ ቇ ⎇ ⓥ ޔVol.35 ޔNo.1 ޔ pp.86-89 ޔ1987. [5] 㜞 ᩮ “ ޔ ⚂ ઃ ߈ ਥ ᚑ ಽ ಽ ᨆ ᴺ ᦺ ”ޔ ୖ ᦠ ᐫ ޔ1995.. 6. ⻢ ㄉ 䈭 䈍 ᧄ ⎇ ⓥ 䈱 ㆀ ⴕ 䈮 䈅 䈢䉍 䇮 ᣣ ᧄ ㅢ ᩣ ᑼ ળ ␠ Ლ 䈎 䉌䉺䉪䉲䊷䈱䊒䊨䊷䊑䊂䊷䉺䉕䈗 ឭ ଏ 䈇䈢䈣䈐䉁䈚䈢䇯䈖䈖䈮ᷓ ⻢ 䈇䈢䈚䉁䈜䇯. ෳ ⠨ ᢥ ₂ [1] T. Fushiki, et al., “Study on Density of Probe Cars Sufficient for Both Level of Area Coverage and Traffic Information Update Cycle,” Proc. of 11th World Congress on ITS Nagoya, CD-ROM, Japan, Oct. 2004. [2] M.Kumagai,et al.,“Spatial Interpolation −30− 8.
(11)
図
関連したドキュメント
Furthermore, we also consider the viscosity shrinking projection method for finding a common element of the set of solutions of the generalized equilibrium problem and the set of
Key words: Best approximation, Best coapproximation, Chebyshev space, Cometric projection, Interpolation, Metric projection, Selection and Weak Cheby- shev
To derive a weak formulation of (1.1)–(1.8), we first assume that the functions v, p, θ and c are a classical solution of our problem. 33]) and substitute the Neumann boundary
If two Banach spaces are completions of a given normed space, then we can use Theorem 3.1 to construct a lin- ear norm-preserving bijection between them, so the completion of a
We study the local dimension of the invariant measure for K for special values of β and use the projection to obtain results on the local dimension of the Bernoulli
Shi, “The essential norm of a composition operator on the Bloch space in polydiscs,” Chinese Journal of Contemporary Mathematics, vol. Chen, “Weighted composition operators from Fp,
Variational iteration method is a powerful and efficient technique in finding exact and approximate solutions for one-dimensional fractional hyperbolic partial differential equations..
This paper presents an investigation into the mechanics of this specific problem and develops an analytical approach that accounts for the effects of geometrical and material data on
