Upgrade to Pro — share decks privately, control downloads, hide ads and more …

A weighting local search algorithm for large-sc...

Shunji Umetani
October 30, 2021

A weighting local search algorithm for large-scale binary integer programs

産業や学術の幅広い分野に現れる組合せ最適化問題に迅速に対応するためには,汎用性と高性能を兼ね備えた数理最適化ソルバーが欠かせません.ところが,汎用的な数理最適化ソルバーでは個々の問題の持つ特徴を活用できないため苦手な問題が少なくないのが現状で,汎用性と高性能のトレードオフを解決することは容易ではありません.本発表では,メタヒューリスティクスの基本的な手法である局所探索法にデータマイニングの手法を採り入れることで,入力データが持つ特徴をアルゴリズムの性能向上に利用するアプローチを提案します.

Shunji Umetani

October 30, 2021
Tweet

More Decks by Shunji Umetani

Other Decks in Research

Transcript

  1. ⾃⼰紹介 • 梅⾕ 俊治(うめたに しゅんじ) • 所属︓⼤阪⼤学 ⼤学院情報科学研究科 数理最適化寄附講座 •

    専⾨分野︓数理最適化(組合せ最適化),アルゴリズム,離散数学など • 現在の主な研究テーマ︓ ü ⼤規模かつ汎⽤的な組合せ最適化問題に対するアルゴリズムの開発 ü 問題構造の解析に基づく組合せ最適化アルゴリズムの開発 ü 数理最適化モデルとアルゴリズムの現実問題への応⽤ • これまでに取り組んだ応⽤事例︓ ü 電気⾃動⾞の充放電計画 ü 対訳⽂の対応付け ü 紙パイプの切出し計画 ü ⾐服の型紙の配置 ü 無⼈搬送⾞の運⾏計画 ü ⼈の移動履歴の推定 2 ü クーポンの配信計画 ü ホテル予約システムの表⽰順 ü カタログのレイアウト計画 ü ⾃動⾞船の配船・運航計画 ü ⾃動⾞船の積付け計画 ü その他いろいろ取り組み中
  2. 「しっかり学ぶ数理最適化」 • 2020年10⽉に「しっかり学ぶ数理最適化」(講談社)を出版. • 1章︓数理最適化⼊⾨ • 2章︓線形計画 ü 線形計画問題の定式化,単体法, 緩和問題と双対定理

    • 3章︓⾮線形計画 ü ⾮線形計画問題の定式化,制約なし最適化問題, 制約つき最適化問題 • 4章︓整数計画と組合せ最適化 ü 整数計画問題の定式化, アルゴリズムの性能と問題の難しさの評価, 効率的に解ける組合せ最適化問題, 分枝限定法と切除平⾯法,近似解法, 局所探索法,メタヒューリスティクス • 演習問題と解答 3 最適化問題の定式化からアルゴリズムまで幅広い内容を解説
  3. 数理最適化による問題解決 • ビッグデータが取得できた︕ビッグデータが分析できた︕ところで, 分析結果をどのように施策に活⽤すれば良い︖ • 数理最適化は,実際の事例より⽣じる多様な制約の下で,意思決定や 計画策定を実現するための有効な⼿段の⼀つ. • 最適化問題︓制約条件を満たす解の中で⽬的関数を最⼩(最⼤)にする 解を求める問題.

    4 ⼈気商品に偏ってない︖ 予算内に収められる︖ 多様な制約を満たせる︖ AIと⼀⼝には ⾔うものの ビッグデータ 機械学習 数理最適化 データの取得 データの分析 意思決定・計画策定 数百万の顧客に推薦する 商品の割当てを求める 超⼤規模な最適化問題 機械学習で個別の割当の 期待利得は推定できるが クーポン配信の事例
  4. 現実問題への数理最適化の適⽤ • 現実問題が既知の最適化問題に⼀致することはまれ. • 論⽂発表されたアルゴリズムの実装が公開されていることはまれ. • 汎⽤の数理最適化ソルバーを利⽤ ü 整数計画問題に定式化して汎⽤の数理最適化ソルバーを適⽤ ü

    問題の特徴を利⽤できないため苦⼿な問題も少なくない • 専⽤のアルゴリズムを開発 ü 問題の特徴を利⽤した⾼性能なアルゴリズムを開発 ü 適⽤範囲が限られる,開発に⼗分な技術と⼿間が必要 6 多様な問題に適⽤可能な 汎⽤性の⾼いアルゴリズム 個々の問題の特徴を利⽤した ⾼性能なアルゴリズム a 整数計画問題 分枝限定法 b c d e 現実世界 a b c d e 現実世界 問題 a 問題 b 問題 e アルゴリズム a アルゴリズム b アルゴリズム e 「汎⽤的」かつ「⾼性能」なソルバーの実現は困難
  5. データを利⽤した組合せ最適化⼿法 • 充⾜可能性問題や整数計画問題は汎⽤的すぎてアルゴリズム性能 の向上に利⽤できる構造がない︖ • 個々の現実問題の⼊⼒データは特徴的な構造を持つ部分の組み合 わせとなる場合がほとんど. • ⼊⼒データの特徴を事前知識として利⽤できれば,理論的に計算 困難な問題でも実験的には優れたアルゴリズムを開発できる.

    8 データ(instance) 問題(problem) 問題に良い構造が ないから難しいよ︕ min cT x s.t. Ax b x 2 {0, 1}n NP困難問題だしなあ 任意のデータを考えればNP困難 だけど,与えられたデータの問題 例が難しいとは限らないですよ x1 x98 x809 x701 x141 x765 x749 x784 x365 x662 x2 x186 x810 x99 x303 x275 x291 x247 x204 x766 x825 x750 x785 x142 x663 x148 x702 x241 x1040 x417 x3 x100 x811 x304 x187 x78 x143 x703 x891 x767 x826 x751 x786 x418 x657 x205 x292 x276 x995 x391 x166 x682 x4 x79 x491 x85 x1064 x101 x658 x167 x188 x392 x833 x426 x368 x463 x775 x10 x398 x689 x173 x382 x160 x758 x1186 x144 x704 x892 x665 x5 x102 x813 x705 x492 x369 x893 x828 x769 x753 x788 x145 x659 x427 x834 x997 x6 x493 x1066 x731 x208 x777 x428 x162 x760 x685 x394 x7 x494 x309 x104 x86 x370 x429 x475 x424 x707 x395 x170 x11 x690 x174 x8 x495 x87 x105 x708 x83 x430 x371 x476 x642 x1127 x12 x175 x668 x691 x896 x310 x171 x396 x461 x192 x9 x496 x311 x106 x193 x84 x431 x372 x477 x643 x267 x381 x709 x1014 x254 x283 x298 x210 x664 x688 x397 x172 x194 x497 x735 x284 x211 x958 x299 x432 x763 x561 x108 x821 x711 x373 x836 x761 x779 x1190 x1070 x196 x388 x764 x287 x301 x212 x499 x13 x500 x274 x416 x435 x185 x900 x14 x109 x318 x712 x1151 x901 x260 x601 x437 x15 x199 x502 x110 x319 x375 x866 x1115 x242 x153 x713 x261 x602 x438 x16 x827 x111 x714 x277 x768 x752 x1179 x787 x200 x320 x376 x671 x293 x903 x1153 x17 x112 x1074 x201 x155 x715 x278 x504 x1180 x868 x244 x1117 x18 x1075 x113 x202 x829 x420 x279 x716 x322 x378 x869 x245 x1118 x754 x770 x789 x19 x203 x114 x487 x1119 x830 x20 x831 x323 x281 x1077 x756 x816 x773 x790 x191 x907 x972 x507 x772 x21 x115 x508 x1078 x324 x717 x154 x56 x282 x444 x832 x871 x22 x116 x509 x718 x325 x1079 x1185 x818 x792 x672 x675 x445 x1030 x425 x774 x23 x117 x510 x326 x1080 x676 x719 x1142 x873 x1143 x446 x1035 x1159 x272 x910 x58 x24 x118 x1081 x835 x157 x383 x720 x776 x1187 x1188 x251 x1125 x874 x25 x209 x286 x513 x253 x876 x794 x300 x195 x912 x448 x1051 x26 x514 x119 x329 x449 x385 x1146 x722 x837 x450 x877 x1147 x1037 x913 x1161 x27 x723 x120 x330 x288 x386 x679 x677 x1148 x451 x1019 x914 x1162 x467 x28 x289 x331 x915 x948 x453 x1150 x839 x29 x121 x290 x726 x686 x389 x30 x122 x846 x520 x727 x339 x390 x994 x917 x455 x1164 x647 x681 x687 x31 x847 x728 x1091 x123 x221 x408 x1158 x791 x803 x521 x32 x124 x222 x522 x1092 x729 x1166 x919 x457 x584 x33 x125 x730 x523 x342 x223 x684 x393 x458 x920 x1160 x1016 x464 x849 x36 x527 x128 x226 x345 x734 x588 x851 x654 x807 x37 x129 x736 x1099 x227 x926 x1173 x399 x38 x1102 x130 x532 x214 x797 x738 x402 x470 x39 x739 x1103 x853 x131 x176 x692 x403 x215 x533 x798 x348 x40 x534 x740 x132 x349 x854 x472 x404 x799 x693 x230 x367 x471 x1165 x41 x741 x1105 x350 x855 x133 x694 x405 x217 x535 x800 x231 x42 x134 x856 x351 x742 x179 x1167 x232 x473 x294 x695 x43 x135 x858 x353 x743 x538 x234 x1169 x802 x1170 x44 x744 x136 x1109 x354 x539 x235 x178 x181 x804 x1000 x1034 x859 x409 x1171 x1048 x596 x45 x355 x137 x860 x540 x236 x1110 x182 x745 x974 x410 x478 x1172 x1015 x1049 x805 x46 x138 x861 x356 x746 x237 x183 x699 x696 x47 x238 x139 x862 x543 x1112 x413 x357 x747 x806 x480 x1174 x48 x748 x140 x863 x358 x239 x1038 x49 x302 x481 x50 x1072 x656 x77 x51 x360 x483 x1041 x629 x482 x1073 x52 x419 x305 x884 x484 x503 x812 x53 x362 x631 x990 x1043 x306 x485 x54 x146 x263 x421 x660 x1137 x886 x307 x486 x1076 x55 x771 x264 x755 x1138 x887 x147 x265 x889 x1139 x488 x489 x1012 x57 x266 x890 x366 x619 x490 x248 x606 x1140 x149 x249 x1141 x340 x554 x59 x150 x268 x312 x250 x608 x206 x1123 x555 x60 x1082 x269 x1124 x1031 x61 x151 x270 x667 x895 x1145 x641 x314 x62 x152 x1085 x1018 x315 x271 x515 x1191 x780 x63 x781 x1086 x838 x560 x64 x273 x434 x257 x1149 x899 x1054 x628 x65 x1055 x442 x246 x374 x1120 x423 x308 x66 x443 x1056 x670 x67 x1057 x1122 x607 x1042 x68 x1058 x377 x872 x69 x1059 x156 x207 x673 x70 x447 x379 x674 x327 x512 x71 x1062 x252 x380 x1126 x313 x1013 x72 x1063 x159 x757 x328 x73 x1128 x567 x74 x161 x759 x255 x878 x819 x1065 x1129 x1050 x511 x568 x75 x1067 x256 x571 x1130 x945 x76 x163 x1068 x1131 x518 x762 x165 x454 x332 x519 x883 x683 x80 x168 x189 x81 x169 x190 x363 x336 x82 x296 x460 x127 x526 x297 x1001 x528 x341 x572 x732 x344 x400 x733 x737 x88 x407 x967 x1028 x537 x1046 x89 x177 x595 x90 x583 x969 x848 x91 x92 x180 x411 x474 x93 x412 x697 x542 x1017 x587 x94 x698 x598 x795 x95 x414 x544 x911 x96 x415 x700 x184 x545 x97 x888 x462 x904 x937 x953 x905 x971 x954 x103 x706 x894 x666 x909 x466 x897 x107 x710 x898 x669 x498 x865 x1177 x867 x1116 x243 x441 x505 x793 x880 x220 x582 x585 x651 x126 x34 x343 x525 x586 x999 x465 x35 x850 x1020 x347 x1006 x1008 x592 x593 x844 x1135 x636 x918 x935 x1045 x506 x1011 x158 x1061 x611 x817 x1053 x678 x384 x724 x721 x164 x680 x725 x1069 x334 x959 x530 x1121 x870 x1029 x1152 x902 x1154 x1009 x1095 x922 x338 x908 x1157 x923 x559 x941 x973 x924 x612 x942 x626 x985 x1036 x925 x627 x927 x977 x563 x197 x864 x1113 x240 x198 x1114 x600 x259 x603 x262 x422 x618 x213 x576 x840 x987 x961 x796 x882 x1021 x841 x333 x1007 x1023 x578 x989 x952 x216 x335 x1104 x1025 x1010 x1155 x991 x965 x1044 x218 x801 x1026 x845 x1106 x857 x604 x992 x219 x993 x557 x640 x594 x982 x224 x652 x570 x983 x597 x1003 x976 x228 x620 x649 x980 x621 x650 x981 x996 x1047 x1052 x625 x653 x984 x614 x1002 x613 x479 x1004 x258 x1132 x1133 x1134 x885 x1136 x949 x591 x565 x951 x955 x562 x929 x1022 x551 x934 x564 x1027 x637 x968 x566 x556 x639 x950 x843 x440 x921 x280 x906 x1156 x337 x998 x459 x558 x940 x1033 x285 x944 x529 x574 x575 x978 x590 x979 x229 x589 x646 x946 x1005 x1039 x814 x524 x617 x623 x316 x433 x823 x782 x1193 x317 x436 x501 x546 x928 x630 x547 x548 x930 x321 x516 x573 x783 x1194 x456 x957 x346 x468 x517 x469 x966 x552 x635 x352 x1168 x938 x406 x295 x359 x361 x577 x579 x1024 x364 x553 x581 x605 x638 x936 x439 x622 x387 x879 x824 x970 x1098 x401 x531 x648 x536 x624 x1111 x956 x655 x986 x549 x632 x931 x550 x633 x932 x1184 x822 x452 x615 x1087 x960 x1178 x943 x644 x975 x1032 x233 x541 x225 x962 x963 x964 x580 x634 x933 x645 x569 x875 x599 x916 x609 x610 x939 x616 x947 x661 x1060 x1084 x1192 x1088 x815 x1183 x778 x820 x1083 x1189 x1090 x842 x1093 x1094 x1096 x1097 x1100 x1101 x852 x1163 x1108 x808 x1175 x881 x988 x1071 x1089 x1107 x1144 x1176 x1181 x1182 データに潜む構造を使えば 上⼿く解けるんじゃない︖ 補助データ(meta data)
  6. 局所探索法 • 適当な初期解 から始めて,現在の解 の近傍 に改善解 があれば移動する(近傍探索). • 勾配法では微分 を⽤いるが,局所探索法では差分

    を⽤いて探索⽅向を決定する. • 0-1整数計画問題の近傍の例 ü 1-flip近傍︓ ü 2-flip近傍︓ 11 初期解 局所最適解 近傍 初期解の⽣成 近傍解の⽣成と評価 近傍内に改善解あり︖ 改善解に移動 局所最適解 yes no
  7. 探索空間と評価関数 • 実⾏可能解を1つ⾒つけることが⾮常に難しい問題も少なくない. • 制約条件を緩和して実⾏不可能解も探索空間に含める. • 各制約 に対する違反度 に重み を掛けた値をペナルティとして

    ⽬的関数に加える. • ペナルティ重み の調整は容易ではなく,適当な値を与えて局所探 索法を1回適⽤するだけでは質の⾼い実⾏可能解は得られない. 12
  8. ペナルティ重みの⾃動調整 • ペナルティ重み の更新と局所探索法を交互に繰り返し適⽤する. • ならば下式にしたがってペナルティ重みを更新,そうで なければ, とする. 13 ペナルティ重みが⼤きい

    実⾏不可能領域 実⾏不可能領域 実⾏不可能領域 最 ⼩ 化 最 ⼩ 化 最 ⼩ 化 ペナルティ重みが⼩さい ペナルティ重みが適切 実⾏不可能領域 実⾏可能領域 有望な解が含まれる領域
  9. 離散構造を利⽤した局所探索法 • 離散構造は特徴的な制約式の組み合わせで表される. ü 割当制約,ナップサック制約,フロー制約など • 専⽤解法の基本操作は整数計画問題では多くの変数を同時に反転 する操作になる. • k個の変数を同時に反転して得られる解候補の数はO(nk)個なので,

    全ての解候補を⾛査するのは効率が悪い. 14 x11 +x12 +・・・+x1n =1 4つの0-1変数を 同時にフリップ 割当制約に対する交換近傍操作 x11 x12 x22 x21 x21 +x22 +・・・+x2n =1 x11 +x12 +・・・+x1n =1 x11 x12 x22 x21 x21 +x22 +・・・+x2n =1
  10. 離散構造を利⽤した局所探索法 • 離散構造は特徴的な制約式の組み合わせで表される. ü 割当制約,ナップサック制約,フロー制約など • 専⽤解法の基本操作は整数計画問題では多くの変数を同時に反転 する操作になる. • k個の変数を同時に反転して得られる解候補の数はO(nk)個なので,

    全ての解候補を⾛査するのは効率が悪い. 15 2つの0-1変数を 同時にフリップ X i xik = X j xkj xik xkj j k xkj' xkj i j k フロー制約に対する近傍操作 xik xkj 4つの0-1変数を 同時にフリップ i i j k kʼ jʼ xik' xk'j
  11. 近傍グラフによる探索空間の縮⼩ • 同時に反転すると改善解が得られる可能性の⾼い変数の組み合わせを 効率良く⾒付けたい. • 同じ制約に同時に現れる頻度が⾼い2つの変数を同時に反転すると改 善解が得られ易い. • ⼊⼒データから特殊な制約に同時に現れる変数の組み合わせを解析し てネットワークを⽣成する

    ← データマイニングにおける近傍探索* 17 xj1 A b c xj2 同じ制約条件に同時に 現れる頻度が⾼い 変数間の関係を表すk-近傍グラフ x1 x98 x809 x701 x141 x765 x749 x784 x365 x662 x2 x186 x810 x99 x303 x275 x291 x247 x204 x766 x825 x750 x785 x142 x663 x148 x702 x241 x1040 x417 x3 x100 x811 x304 x187 x78 x143 x703 x891 x767 x826 x751 x786 x418 x657 x205 x292 x276 x995 x391 x166 x682 x4 x79 x491 x85 x1064 x101 x658 x167 x188 x392 x833 x426 x368 x463 x775 x10 x398 x689 x173 x382 x160 x758 x1186 x144 x704 x892 x665 x5 x102 x813 x705 x492 x369 x893 x828 x769 x753 x788 x145 x659 x427 x834 x997 x6 x493 x1066 x731 x208 x777 x428 x162 x760 x685 x394 x7 x494 x309 x104 x86 x370 x429 x475 x424 x707 x395 x170 x11 x690 x174 x8 x495 x87 x105 x708 x83 x430 x371 x476 x642 x1127 x12 x175 x668 x691 x896 x310 x171 x396 x461 x192 x9 x496 x311 x106 x193 x84 x431 x372 x477 x643 x267 x381 x709 x1014 x254 x283 x298 x210 x664 x688 x397 x172 x194 x497 x735 x284 x211 x958 x299 x432 x763 x561 x108 x821 x711 x373 x836 x761 x779 x1190 x1070 x196 x388 x764 x287 x301 x212 x499 x13 x500 x274 x416 x435 x185 x900 x14 x109 x318 x712 x1151 x901 x260 x601 x437 x15 x199 x502 x110 x319 x375 x866 x1115 x242 x153 x713 x261 x602 x438 x16 x827 x111 x714 x277 x768 x752 x1179 x787 x200 x320 x376 x671 x293 x903 x1153 x17 x112 x1074 x201 x155 x715 x278 x504 x1180 x868 x244 x1117 x18 x1075 x113 x202 x829 x420 x279 x716 x322 x378 x869 x245 x1118 x754 x770 x789 x19 x203 x114 x487 x1119 x830 x20 x831 x323 x281 x1077 x756 x816 x773 x790 x191 x907 x972 x507 x772 x21 x115 x508 x1078 x324 x717 x154 x56 x282 x444 x832 x871 x22 x116 x509 x718 x325 x1079 x1185 x818 x792 x672 x675 x445 x1030 x425 x774 x23 x117 x510 x326 x1080 x676 x719 x1142 x873 x1143 x446 x1035 x1159 x272 x910 x58 x24 x118 x1081 x835 x157 x383 x720 x776 x1187 x1188 x251 x1125 x874 x25 x209 x286 x513 x253 x876 x794 x300 x195 x912 x448 x1051 x26 x514 x119 x329 x449 x385 x1146 x722 x837 x450 x877 x1147 x1037 x913 x1161 x27 x723 x120 x330 x288 x386 x679 x677 x1148 x451 x1019 x914 x1162 x467 x28 x289 x331 x915 x948 x453 x1150 x839 x29 x121 x290 x726 x686 x389 x30 x122 x846 x520 x727 x339 x390 x994 x917 x455 x1164 x647 x681 x687 x31 x847 x728 x1091 x123 x221 x408 x1158 x791 x803 x521 x32 x124 x222 x522 x1092 x729 x1166 x919 x457 x584 x33 x125 x730 x523 x342 x223 x684 x393 x458 x920 x1160 x1016 x464 x849 x36 x527 x128 x226 x345 x734 x588 x851 x654 x807 x37 x129 x736 x1099 x227 x926 x1173 x399 x38 x1102 x130 x532 x214 x797 x738 x402 x470 x39 x739 x1103 x853 x131 x176 x692 x403 x215 x533 x798 x348 x40 x534 x740 x132 x349 x854 x472 x404 x799 x693 x230 x367 x471 x1165 x41 x741 x1105 x350 x855 x133 x694 x405 x217 x535 x800 x231 x42 x134 x856 x351 x742 x179 x1167 x232 x473 x294 x695 x43 x135 x858 x353 x743 x538 x234 x1169 x802 x1170 x44 x744 x136 x1109 x354 x539 x235 x178 x181 x804 x1000 x1034 x859 x409 x1171 x1048 x596 x45 x355 x137 x860 x540 x236 x1110 x182 x745 x974 x410 x478 x1172 x1015 x1049 x805 x46 x138 x861 x356 x746 x237 x183 x699 x696 x47 x238 x139 x862 x543 x1112 x413 x357 x747 x806 x480 x1174 x48 x748 x140 x863 x358 x239 x1038 x49 x302 x481 x50 x1072 x656 x77 x51 x360 x483 x1041 x629 x482 x1073 x52 x419 x305 x884 x484 x503 x812 x53 x362 x631 x990 x1043 x306 x485 x54 x146 x263 x421 x660 x1137 x886 x307 x486 x1076 x55 x771 x264 x755 x1138 x887 x147 x265 x889 x1139 x488 x489 x1012 x57 x266 x890 x366 x619 x490 x248 x606 x1140 x149 x249 x1141 x340 x554 x59 x150 x268 x312 x250 x608 x206 x1123 x555 x60 x1082 x269 x1124 x1031 x61 x151 x270 x667 x895 x1145 x641 x314 x62 x152 x1085 x1018 x315 x271 x515 x1191 x780 x63 x781 x1086 x838 x560 x64 x273 x434 x257 x1149 x899 x1054 x628 x65 x1055 x442 x246 x374 x1120 x423 x308 x66 x443 x1056 x670 x67 x1057 x1122 x607 x1042 x68 x1058 x377 x872 x69 x1059 x156 x207 x673 x70 x447 x379 x674 x327 x512 x71 x1062 x252 x380 x1126 x313 x1013 x72 x1063 x159 x757 x328 x73 x1128 x567 x74 x161 x759 x255 x878 x819 x1065 x1129 x1050 x511 x568 x75 x1067 x256 x571 x1130 x945 x76 x163 x1068 x1131 x518 x762 x165 x454 x332 x519 x883 x683 x80 x168 x189 x81 x169 x190 x363 x336 x82 x296 x460 x127 x526 x297 x1001 x528 x341 x572 x732 x344 x400 x733 x737 x88 x407 x967 x1028 x537 x1046 x89 x177 x595 x90 x583 x969 x848 x91 x92 x180 x411 x474 x93 x412 x697 x542 x1017 x587 x94 x698 x598 x795 x95 x414 x544 x911 x96 x415 x700 x184 x545 x97 x888 x462 x904 x937 x953 x905 x971 x954 x103 x706 x894 x666 x909 x466 x897 x107 x710 x898 x669 x498 x865 x1177 x867 x1116 x243 x441 x505 x793 x880 x220 x582 x585 x651 x126 x34 x343 x525 x586 x999 x465 x35 x850 x1020 x347 x1006 x1008 x592 x593 x844 x1135 x636 x918 x935 x1045 x506 x1011 x158 x1061 x611 x817 x1053 x678 x384 x724 x721 x164 x680 x725 x1069 x334 x959 x530 x1121 x870 x1029 x1152 x902 x1154 x1009 x1095 x922 x338 x908 x1157 x923 x559 x941 x973 x924 x612 x942 x626 x985 x1036 x925 x627 x927 x977 x563 x197 x864 x1113 x240 x198 x1114 x600 x259 x603 x262 x422 x618 x213 x576 x840 x987 x961 x796 x882 x1021 x841 x333 x1007 x1023 x578 x989 x952 x216 x335 x1104 x1025 x1010 x1155 x991 x965 x1044 x218 x801 x1026 x845 x1106 x857 x604 x992 x219 x993 x557 x640 x594 x982 x224 x652 x570 x983 x597 x1003 x976 x228 x620 x649 x980 x621 x650 x981 x996 x1047 x1052 x625 x653 x984 x614 x1002 x613 x479 x1004 x258 x1132 x1133 x1134 x885 x1136 x949 x591 x565 x951 x955 x562 x929 x1022 x551 x934 x564 x1027 x637 x968 x566 x556 x639 x950 x843 x440 x921 x280 x906 x1156 x337 x998 x459 x558 x940 x1033 x285 x944 x529 x574 x575 x978 x590 x979 x229 x589 x646 x946 x1005 x1039 x814 x524 x617 x623 x316 x433 x823 x782 x1193 x317 x436 x501 x546 x928 x630 x547 x548 x930 x321 x516 x573 x783 x1194 x456 x957 x346 x468 x517 x469 x966 x552 x635 x352 x1168 x938 x406 x295 x359 x361 x577 x579 x1024 x364 x553 x581 x605 x638 x936 x439 x622 x387 x879 x824 x970 x1098 x401 x531 x648 x536 x624 x1111 x956 x655 x986 x549 x632 x931 x550 x633 x932 x1184 x822 x452 x615 x1087 x960 x1178 x943 x644 x975 x1032 x233 x541 x225 x962 x963 x964 x580 x634 x933 x645 x569 x875 x599 x916 x609 x610 x939 x616 x947 x661 x1060 x1084 x1192 x1088 x815 x1183 x778 x820 x1083 x1189 x1090 x842 x1093 x1094 x1096 x1097 x1100 x1101 x852 x1163 x1108 x808 x1175 x881 x988 x1071 x1089 x1107 x1144 x1176 x1181 x1182 *この「近傍探索」は機械学習における類似データの探索を指す.
  12. 近傍グラフによる探索の集中化 • 各変数xj について同じ制約に同時に現れる変数を列挙. • ivu06-big(約227万変数)など変数が⾮常に多い問題例は少なくない. • 組合せの数がn2(nは変数の数)となるので,xj と同時に現れる頻度の ⾼い変数のみを残して近傍グラフを構成.

    • 多くの計算時間を要するので,変数xj を反転する際に近傍グラフの必 要な部分のみ遅延⽣成する ← 実際に利⽤されるのは⼀部分だけ 18 x1 x98 x809 x701 x141 x765 x749 x784 x365 x662 x2 x186 x810 x99 x303 x275 x291 x247 x204 x766 x825 x750 x785 x142 x663 x148 x702 x241 x1040 x417 x3 x100 x811 x304 x187 x78 x143 x703 x891 x767 x826 x751 x786 x418 x657 x205 x292 x276 x995 x391 x166 x682 x4 x79 x491 x85 x1064 x101 x658 x167 x188 x392 x833 x426 x368 x463 x775 x10 x398 x689 x173 x382 x160 x758 x1186 x144 x704 x892 x665 x5 x102 x813 x705 x492 x369 x893 x828 x769 x753 x788 x145 x659 x427 x834 x997 x6 x493 x1066 x731 x208 x777 x428 x162 x760 x685 x394 x7 x494 x309 x104 x86 x370 x429 x475 x424 x707 x395 x170 x11 x690 x174 x8 x495 x87 x105 x708 x83 x430 x371 x476 x642 x1127 x12 x175 x668 x691 x896 x310 x171 x396 x461 x192 x9 x496 x311 x106 x193 x84 x431 x372 x477 x643 x267 x381 x709 x1014 x254 x283 x298 x210 x664 x688 x397 x172 x194 x497 x735 x284 x211 x958 x299 x432 x763 x561 x108 x821 x711 x373 x836 x761 x779 x1190 x1070 x196 x388 x764 x287 x301 x212 x499 x13 x500 x274 x416 x435 x185 x900 x14 x109 x318 x712 x1151 x901 x260 x601 x437 x15 x199 x502 x110 x319 x375 x866 x1115 x242 x153 x713 x261 x602 x438 x16 x827 x111 x714 x277 x768 x752 x1179 x787 x200 x320 x376 x671 x293 x903 x1153 x17 x112 x1074 x201 x155 x715 x278 x504 x1180 x868 x244 x1117 x18 x1075 x113 x202 x829 x420 x279 x716 x322 x378 x869 x245 x1118 x754 x770 x789 x19 x203 x114 x487 x1119 x830 x20 x831 x323 x281 x1077 x756 x816 x773 x790 x191 x907 x972 x507 x772 x21 x115 x508 x1078 x324 x717 x154 x56 x282 x444 x832 x871 x22 x116 x509 x718 x325 x1079 x1185 x818 x792 x672 x675 x445 x1030 x425 x774 x23 x117 x510 x326 x1080 x676 x719 x1142 x873 x1143 x446 x1035 x1159 x272 x910 x58 x24 x118 x1081 x835 x157 x383 x720 x776 x1187 x1188 x251 x1125 x874 x25 x209 x286 x513 x253 x876 x794 x300 x195 x912 x448 x1051 x26 x514 x119 x329 x449 x385 x1146 x722 x837 x450 x877 x1147 x1037 x913 x1161 x27 x723 x120 x330 x288 x386 x679 x677 x1148 x451 x1019 x914 x1162 x467 x28 x289 x331 x915 x948 x453 x1150 x839 x29 x121 x290 x726 x686 x389 x30 x122 x846 x520 x727 x339 x390 x994 x917 x455 x1164 x647 x681 x687 x31 x847 x728 x1091 x123 x221 x408 x1158 x791 x803 x521 x32 x124 x222 x522 x1092 x729 x1166 x919 x457 x584 x33 x125 x730 x523 x342 x223 x684 x393 x458 x920 x1160 x1016 x464 x849 x36 x527 x128 x226 x345 x734 x588 x851 x654 x807 x37 x129 x736 x1099 x227 x926 x1173 x399 x38 x1102 x130 x532 x214 x797 x738 x402 x470 x39 x739 x1103 x853 x131 x176 x692 x403 x215 x533 x798 x348 x40 x534 x740 x132 x349 x854 x472 x404 x799 x693 x230 x367 x471 x1165 x41 x741 x1105 x350 x855 x133 x694 x405 x217 x535 x800 x231 x42 x134 x856 x351 x742 x179 x1167 x232 x473 x294 x695 x43 x135 x858 x353 x743 x538 x234 x1169 x802 x1170 x44 x744 x136 x1109 x354 x539 x235 x178 x181 x804 x1000 x1034 x859 x409 x1171 x1048 x596 x45 x355 x137 x860 x540 x236 x1110 x182 x745 x974 x410 x478 x1172 x1015 x1049 x805 x46 x138 x861 x356 x746 x237 x183 x699 x696 x47 x238 x139 x862 x543 x1112 x413 x357 x747 x806 x480 x1174 x48 x748 x140 x863 x358 x239 x1038 x49 x302 x481 x50 x1072 x656 x77 x51 x360 x483 x1041 x629 x482 x1073 x52 x419 x305 x884 x484 x503 x812 x53 x362 x631 x990 x1043 x306 x485 x54 x146 x263 x421 x660 x1137 x886 x307 x486 x1076 x55 x771 x264 x755 x1138 x887 x147 x265 x889 x1139 x488 x489 x1012 x57 x266 x890 x366 x619 x490 x248 x606 x1140 x149 x249 x1141 x340 x554 x59 x150 x268 x312 x250 x608 x206 x1123 x555 x60 x1082 x269 x1124 x1031 x61 x151 x270 x667 x895 x1145 x641 x314 x62 x152 x1085 x1018 x315 x271 x515 x1191 x780 x63 x781 x1086 x838 x560 x64 x273 x434 x257 x1149 x899 x1054 x628 x65 x1055 x442 x246 x374 x1120 x423 x308 x66 x443 x1056 x670 x67 x1057 x1122 x607 x1042 x68 x1058 x377 x872 x69 x1059 x156 x207 x673 x70 x447 x379 x674 x327 x512 x71 x1062 x252 x380 x1126 x313 x1013 x72 x1063 x159 x757 x328 x73 x1128 x567 x74 x161 x759 x255 x878 x819 x1065 x1129 x1050 x511 x568 x75 x1067 x256 x571 x1130 x945 x76 x163 x1068 x1131 x518 x762 x165 x454 x332 x519 x883 x683 x80 x168 x189 x81 x169 x190 x363 x336 x82 x296 x460 x127 x526 x297 x1001 x528 x341 x572 x732 x344 x400 x733 x737 x88 x407 x967 x1028 x537 x1046 x89 x177 x595 x90 x583 x969 x848 x91 x92 x180 x411 x474 x93 x412 x697 x542 x1017 x587 x94 x698 x598 x795 x95 x414 x544 x911 x96 x415 x700 x184 x545 x97 x888 x462 x904 x937 x953 x905 x971 x954 x103 x706 x894 x666 x909 x466 x897 x107 x710 x898 x669 x498 x865 x1177 x867 x1116 x243 x441 x505 x793 x880 x220 x582 x585 x651 x126 x34 x343 x525 x586 x999 x465 x35 x850 x1020 x347 x1006 x1008 x592 x593 x844 x1135 x636 x918 x935 x1045 x506 x1011 x158 x1061 x611 x817 x1053 x678 x384 x724 x721 x164 x680 x725 x1069 x334 x959 x530 x1121 x870 x1029 x1152 x902 x1154 x1009 x1095 x922 x338 x908 x1157 x923 x559 x941 x973 x924 x612 x942 x626 x985 x1036 x925 x627 x927 x977 x563 x197 x864 x1113 x240 x198 x1114 x600 x259 x603 x262 x422 x618 x213 x576 x840 x987 x961 x796 x882 x1021 x841 x333 x1007 x1023 x578 x989 x952 x216 x335 x1104 x1025 x1010 x1155 x991 x965 x1044 x218 x801 x1026 x845 x1106 x857 x604 x992 x219 x993 x557 x640 x594 x982 x224 x652 x570 x983 x597 x1003 x976 x228 x620 x649 x980 x621 x650 x981 x996 x1047 x1052 x625 x653 x984 x614 x1002 x613 x479 x1004 x258 x1132 x1133 x1134 x885 x1136 x949 x591 x565 x951 x955 x562 x929 x1022 x551 x934 x564 x1027 x637 x968 x566 x556 x639 x950 x843 x440 x921 x280 x906 x1156 x337 x998 x459 x558 x940 x1033 x285 x944 x529 x574 x575 x978 x590 x979 x229 x589 x646 x946 x1005 x1039 x814 x524 x617 x623 x316 x433 x823 x782 x1193 x317 x436 x501 x546 x928 x630 x547 x548 x930 x321 x516 x573 x783 x1194 x456 x957 x346 x468 x517 x469 x966 x552 x635 x352 x1168 x938 x406 x295 x359 x361 x577 x579 x1024 x364 x553 x581 x605 x638 x936 x439 x622 x387 x879 x824 x970 x1098 x401 x531 x648 x536 x624 x1111 x956 x655 x986 x549 x632 x931 x550 x633 x932 x1184 x822 x452 x615 x1087 x960 x1178 x943 x644 x975 x1032 x233 x541 x225 x962 x963 x964 x580 x634 x933 x645 x569 x875 x599 x916 x609 x610 x939 x616 x947 x661 x1060 x1084 x1192 x1088 x815 x1183 x778 x820 x1083 x1189 x1090 x842 x1093 x1094 x1096 x1097 x1100 x1101 x852 x1163 x1108 x808 x1175 x881 x988 x1071 x1089 x1107 x1144 x1176 x1181 x1182 x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x98 x809 x701 x186 x810 x99 x303 x100 x811 x304 x187 x78 x79 x491 x85 x1064 x101 x102 x813 x705 x492 x493 x1066 x731 x494 x309 x104 x86 x495 x87 x105 x708 x83 x496 x311 x106 x193 x84 x194 x85 x497 x735 x108 x821 x86 同じ割当制約に同時に 現れる変数を列挙 頻度の⾼い変数のみ残す
  13. 近傍グラフを利⽤した⼤規模近傍探索 • 1-flip近傍操作だけでは精度の良い解は得られない. • 変数が⾮常に多い問題例では,多数の変数を同時に反転する近傍 操作は時間がかかり過ぎる. • 2-flip近傍操作はネットワークの辺にある変数の組み合わせに対 してのみ適⽤する. •

    2-flip近傍操作で改善しなかったものの有望な組み合わせを記憶 しておき,これらを組み合わせて(3,4)-flip近傍操作を実現する. 19 x998 x1032 x610 x939 x1 x98 x809 x701 x141 x765 x749 x784 x365 x662 x2 x186 x810 x99 x303 x275 x291 x247 x204 x766 x825 x750 x785 x142 x663 x148 x702 x241 x1040 x417 x3 x100 x811 x304 x187 x78 x143 x703 x891 x767 x826 x751 x786 x418 x657 x205 x292 x276 x995 x391 x166 x682 x4 x79 x491 x85 x1064 x101 x658 x167 x188 x392 x833 x426 x368 x463 x775 x10 x398 x689 x173 x382 x160 x758 x1186 x144 x704 x892 x665 x5 x102 x813 x705 x492 x369 x893 x828 x769 x753 x788 x145 x659 x427 x834 x997 x6 x493 x1066 x731 x208 x777 x428 x162 x760 x685 x394 x7 x494 x309 x104 x86 x370 x429 x475 x424 x707 x395 x170 x11 x690 x174 x8 x495 x87 x105 x708 x83 x430 x371 x476 x642 x1127 x12 x175 x668 x691 x896 x310 x171 x396 x461 x192 x9 x496 x311 x106 x193 x84 x431 x372 x477 x643 x267 x381 x709 x1014 x254 x283 x298 x210 x664 x688 x397 x172 x194 x497 x735 x284 x211 x958 x299 x432 x763 x561 x108 x821 x711 x373 x836 x761 x779 x1190 x1070 x196 x388 x764 x287 x301 x212 x499 x13 x500 x274 x416 x435 x185 x900 x14 x109 x318 x712 x1151 x901 x260 x601 x437 x15 x199 x502 x110 x319 x375 x866 x1115 x242 x153 x713 x261 x602 x438 x16 x827 x111 x714 x277 x768 x752 x1179 x787 x200 x320 x376 x671 x293 x903 x1153 x17 x112 x1074 x201 x155 x715 x278 x504 x1180 x868 x244 x1117 x18 x1075 x113 x202 x829 x420 x279 x716 x322 x378 x869 x245 x1118 x754 x770 x789 x19 x203 x114 x487 x1119 x830 x20 x831 x323 x281 x1077 x756 x816 x773 x790 x191 x907 x972 x507 x772 x21 x115 x508 x1078 x324 x717 x154 x56 x282 x444 x832 x871 x22 x116 x509 x718 x325 x1079 x1185 x818 x792 x672 x675 x445 x1030 x425 x774 x23 x117 x510 x326 x1080 x676 x719 x1142 x873 x1143 x446 x1035 x1159 x272 x910 x58 x24 x118 x1081 x835 x157 x383 x720 x776 x1187 x1188 x251 x1125 x874 x25 x209 x286 x513 x253 x876 x794 x300 x195 x912 x448 x1051 x26 x514 x119 x329 x449 x385 x1146 x722 x837 x450 x877 x1147 x1037 x913 x1161 x27 x723 x120 x330 x288 x386 x679 x677 x1148 x451 x1019 x914 x1162 x467 x28 x289 x331 x915 x948 x453 x1150 x839 x29 x121 x290 x726 x686 x389 x30 x122 x846 x520 x727 x339 x390 x994 x917 x455 x1164 x647 x681 x687 x31 x847 x728 x1091 x123 x221 x408 x1158 x791 x803 x521 x32 x124 x222 x522 x1092 x729 x1166 x919 x457 x584 x33 x125 x730 x523 x342 x223 x684 x393 x458 x920 x1160 x1016 x464 x849 x36 x527 x128 x226 x345 x734 x588 x851 x654 x807 x37 x129 x736 x1099 x227 x926 x1173 x399 x38 x1102 x130 x532 x214 x797 x738 x402 x470 x39 x739 x1103 x853 x131 x176 x692 x403 x215 x533 x798 x348 x40 x534 x740 x132 x349 x854 x472 x404 x799 x693 x230 x367 x471 x1165 x41 x741 x1105 x350 x855 x133 x694 x405 x217 x535 x800 x231 x42 x134 x856 x351 x742 x179 x1167 x232 x473 x294 x695 x43 x135 x858 x353 x743 x538 x234 x1169 x802 x1170 x44 x744 x136 x1109 x354 x539 x235 x178 x181 x804 x1000 x1034 x859 x409 x1171 x1048 x596 x45 x355 x137 x860 x540 x236 x1110 x182 x745 x974 x410 x478 x1172 x1015 x1049 x805 x46 x138 x861 x356 x746 x237 x183 x699 x696 x47 x238 x139 x862 x543 x1112 x413 x357 x747 x806 x480 x1174 x48 x748 x140 x863 x358 x239 x1038 x49 x302 x481 x50 x1072 x656 x77 x51 x360 x483 x1041 x629 x482 x1073 x52 x419 x305 x884 x484 x503 x812 x53 x362 x631 x990 x1043 x306 x485 x54 x146 x263 x421 x660 x1137 x886 x307 x486 x1076 x55 x771 x264 x755 x1138 x887 x147 x265 x889 x1139 x488 x489 x1012 x57 x266 x890 x366 x619 x490 x248 x606 x1140 x149 x249 x1141 x340 x554 x59 x150 x268 x312 x250 x608 x206 x1123 x555 x60 x1082 x269 x1124 x1031 x61 x151 x270 x667 x895 x1145 x641 x314 x62 x152 x1085 x1018 x315 x271 x515 x1191 x780 x63 x781 x1086 x838 x560 x64 x273 x434 x257 x1149 x899 x1054 x628 x65 x1055 x442 x246 x374 x1120 x423 x308 x66 x443 x1056 x670 x67 x1057 x1122 x607 x1042 x68 x1058 x377 x872 x69 x1059 x156 x207 x673 x70 x447 x379 x674 x327 x512 x71 x1062 x252 x380 x1126 x313 x1013 x72 x1063 x159 x757 x328 x73 x1128 x567 x74 x161 x759 x255 x878 x819 x1065 x1129 x1050 x511 x568 x75 x1067 x256 x571 x1130 x945 x76 x163 x1068 x1131 x518 x762 x165 x454 x332 x519 x883 x683 x80 x168 x189 x81 x169 x190 x363 x336 x82 x296 x460 x127 x526 x297 x1001 x528 x341 x572 x732 x344 x400 x733 x737 x88 x407 x967 x1028 x537 x1046 x89 x177 x595 x90 x583 x969 x848 x91 x92 x180 x411 x474 x93 x412 x697 x542 x1017 x587 x94 x698 x598 x795 x95 x414 x544 x911 x96 x415 x700 x184 x545 x97 x888 x462 x904 x937 x953 x905 x971 x954 x103 x706 x894 x666 x909 x466 x897 x107 x710 x898 x669 x498 x865 x1177 x867 x1116 x243 x441 x505 x793 x880 x220 x582 x585 x651 x126 x34 x343 x525 x586 x999 x465 x35 x850 x1020 x347 x1006 x1008 x592 x593 x844 x1135 x636 x918 x935 x1045 x506 x1011 x158 x1061 x611 x817 x1053 x678 x384 x724 x721 x164 x680 x725 x1069 x334 x959 x530 x1121 x870 x1029 x1152 x902 x1154 x1009 x1095 x922 x338 x908 x1157 x923 x559 x941 x973 x924 x612 x942 x626 x985 x1036 x925 x627 x927 x977 x563 x197 x864 x1113 x240 x198 x1114 x600 x259 x603 x262 x422 x618 x213 x576 x840 x987 x961 x796 x882 x1021 x841 x333 x1007 x1023 x578 x989 x952 x216 x335 x1104 x1025 x1010 x1155 x991 x965 x1044 x218 x801 x1026 x845 x1106 x857 x604 x992 x219 x993 x557 x640 x594 x982 x224 x652 x570 x983 x597 x1003 x976 x228 x620 x649 x980 x621 x650 x981 x996 x1047 x1052 x625 x653 x984 x614 x1002 x613 x479 x1004 x258 x1132 x1133 x1134 x885 x1136 x949 x591 x565 x951 x955 x562 x929 x1022 x551 x934 x564 x1027 x637 x968 x566 x556 x639 x950 x843 x440 x921 x280 x906 x1156 x337 x998 x459 x558 x940 x1033 x285 x944 x529 x574 x575 x978 x590 x979 x229 x589 x646 x946 x1005 x1039 x814 x524 x617 x623 x316 x433 x823 x782 x1193 x317 x436 x501 x546 x928 x630 x547 x548 x930 x321 x516 x573 x783 x1194 x456 x957 x346 x468 x517 x469 x966 x552 x635 x352 x1168 x938 x406 x295 x359 x361 x577 x579 x1024 x364 x553 x581 x605 x638 x936 x439 x622 x387 x879 x824 x970 x1098 x401 x531 x648 x536 x624 x1111 x956 x655 x986 x549 x632 x931 x550 x633 x932 x1184 x822 x452 x615 x1087 x960 x1178 x943 x644 x975 x1032 x233 x541 x225 x962 x963 x964 x580 x634 x933 x645 x569 x875 x599 x916 x609 x610 x939 x616 x947 x661 x1060 x1084 x1192 x1088 x815 x1183 x778 x820 x1083 x1189 x1090 x842 x1093 x1094 x1096 x1097 x1100 x1101 x852 x1163 x1108 x808 x1175 x881 x988 x1071 x1089 x1107 x1144 x1176 x1181 x1182 近傍グラフに沿って 順々に変数を反転 ↑ ↑ ↓ ↓
  14. 評価計算の⾼速化 • 近傍操作ではごく少数の変数のみ値が変化するため,現在の解x と近傍解 xʼ ∈NB(x) の間で値が変化した変数に関わる部分のみ 再計算すれば,評価関数値の変化量 Δz(x) =

    z(xʼ) – z(x) を⾼ 速に計算できる場合が多い. • 局所探索法では,近傍内の解を評価する回数に⽐べて,現在の解 が移動する回数ははるかに少ないので,補助記憶の更新に多少時 間がかかっても,全体では⼗分な⾼速化が実現できる. 20 評価関数の変化量 補助記憶 係数⾏列 補助記憶を利⽤すれば 定数時間で計算できる 疎な⾏列だと効率良く 補助記憶が更新できる 本来は⽬的関数と制約条 件の評価に⼊⼒サイズの 計算量が必要 A b c
  15. 数値実験の環境 • ⼤規模な集合被覆問題と集合分割問題について評価. • MacBookPro (Intel Core i7 2.7GHz, 16GBメモリ)

    1スレッドで 実⾏. 21 集合被覆問題のベンチマーク問題例 *前処理によって冗⻑な変数や制約を削除している. 論⽂発表時(2017年)の結果です
  16. 数値実験の環境 • ⼤規模な集合被覆問題と集合分割問題について評価. • MacBookPro (Intel Core i7 2.7GHz, 16GBメモリ)

    1スレッドで 実⾏. 22 集合分割問題のベンチマーク問題例 約256万変数におよぶ ⼤規模な整数計画問題 *前処理によって冗⻑な変数や制約を削除している. 論⽂発表時(2017年)の結果です
  17. 数値実験の結果 • 整数計画ソルバー(CPLEX12.6, Gurobi5.6.3, SCIP3.1, LocalSolver3.1), 専⽤解法(Yagiura et al.)と⽐較. •

    最良値からの相対誤差 (%)で評価. 25 集合被覆問題の実⾏結果 専⽤解法 論⽂発表時(2017年)の結果です
  18. まとめ • 数理最適化による問題解決 → 「分析」から「計画」へ • 現実問題への数理最適化の適⽤ → 汎⽤の数理最適化ソルバーを利⽤したモデリングの効率化 •

    メタヒューリスティクスにもとづく汎⽤ソルバーの開発 → 汎⽤的かつ⾼性能なソルバーの実現 →ペナルティ重みの⾃動調整 → データマイニング⼿法を⽤いた探索の集中化 → 評価計算の⾼速化 → ⼤規模な問題例に対する数値実験 27 現実問題の解決に数理最適化を活⽤して下さい︕
  19. 参考⽂献 • S.Umetani, Exploiting variable associations to configure efficient local

    search algorithm in large-scale set partitioning problems, Proc. of 29th AAAI Conference on Artificial Intelligence (AAAI-15), 1226-1232. • S.Umetani, Exploiting variable associations to configure efficient local search algorithm in large-scale binary integer programs, European Journal of Operational Research, 263 (2017), 72-81. (open access) 28