Lock in $30 Savings on PRO—Offer Ends Soon! ⏳
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Grundy
Search
yosupo
November 18, 2016
Technology
1
3.1k
Grundy
yosupo
November 18, 2016
Tweet
Share
More Decks by yosupo
See All by yosupo
一般グラフの最大マッチング
yosupo
3
2.1k
Other Decks in Technology
See All in Technology
意外と知らない状態遷移テストの世界
nihonbuson
PRO
1
270
事業の財務責任に向き合うリクルートデータプラットフォームのFinOps
recruitengineers
PRO
2
220
フルカイテン株式会社 エンジニア向け採用資料
fullkaiten
0
9.9k
AgentCoreとStrandsで社内d払いナレッジボットを作った話
motojimayu
1
980
「もしもデータ基盤開発で『強くてニューゲーム』ができたなら今の僕はどんなデータ基盤を作っただろう」
aeonpeople
0
250
ActiveJobUpdates
igaiga
1
320
20251218_AIを活用した開発生産性向上の全社的な取り組みの進め方について / How to proceed with company-wide initiatives to improve development productivity using AI
yayoi_dd
0
710
Strands Agents × インタリーブ思考 で変わるAIエージェント設計 / Strands Agents x Interleaved Thinking AI Agents
takanorig
5
2.1k
Next.js 16の新機能 Cache Components について
sutetotanuki
0
190
"人"が頑張るAI駆動開発
yokomachi
1
610
100以上の新規コネクタ提供を可能にしたアーキテクチャ
ooyukioo
0
260
AIエージェント開発と活用を加速するワークフロー自動生成への挑戦
shibuiwilliam
5
860
Featured
See All Featured
Automating Front-end Workflow
addyosmani
1371
200k
Darren the Foodie - Storyboard
khoart
PRO
0
2k
Winning Ecommerce Organic Search in an AI Era - #searchnstuff2025
aleyda
0
1.8k
Design of three-dimensional binary manipulators for pick-and-place task avoiding obstacles (IECON2024)
konakalab
0
320
Context Engineering - Making Every Token Count
addyosmani
9
560
[RailsConf 2023] Rails as a piece of cake
palkan
58
6.2k
How to make the Groovebox
asonas
2
1.8k
Bootstrapping a Software Product
garrettdimon
PRO
307
120k
The untapped power of vector embeddings
frankvandijk
1
1.5k
The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs
inesmontani
PRO
3
2.8k
B2B Lead Gen: Tactics, Traps & Triumph
marketingsoph
0
33
Fashionably flexible responsive web design (full day workshop)
malarkey
407
66k
Transcript
Grundy @yosupot
ҎԼͷΛߟ͑·͢ NMลͷDAG͕༩͑ΒΕΔɻ1ʹ͕ۨஔ͍ͯ͋Δ A͞ΜͱB͞Μ͕ަޓʹۨΛಈ͔͢ɻ ۨลͰͭͳ͕ͬͨʹಈ͔ͤͯɺಈ͔ͤͳ͘ͳͬͨΒ ෛ͚ Ͳ͕ͬͪউͭʁ
ղ๏ dfsͰղ͘ ಈ͔ͤΔʹʮޙʯͱॻ͔Ε͕ͨ͋Εʮઌʯ ͳ͚Εʮޙʯͱॻ͘ 0ʹʮઌʯͱॻ͔ΕΕઌखඞউ ʮޙʯͱॻ͔ΕΕޙखඞউ
ྫ ޙ ޙ ઌ ઌ 1
ҎԼͷΛߟ͑·͢ NMลͷDAG͕Kछྨ༩͑ΒΕΔɻͦΕͧΕͷάϥϑͷ 1ʹ͕ۨஔ͍ͯ͋Δ A͞ΜͱB͞Μ͕ަޓʹɺͲΕ͔ͷάϥϑͷۨΛಈ͔͢ɻ ۨลͰͭͳ͕ͬͨʹಈ͔ͤͯɺಈ͔ͤͳ͘ͳͬͨΒ ෛ͚ Ͳ͕ͬͪউͭʁ
ղ๏ ࣮ෳͷάϥϑ1ͭͷάϥϑʹ·ͱΊΒΕΔ (άϥϑ1ͷۨͷҐஔ, άϥϑ2ͷۨͷҐஔ, …, άϥϑKͷۨͷ Ґஔ)Λͱͯ͠దʹลΛషΔ ͜ͷάϥϑͰಉ͡Α͏ʹdfs͕Ͱ͖Δ ܭࢉྔO((N+M)^K)
O((N+M)^K)ʂʁ N=M=5000ͱͯ͠(N+M)^Kʁ K=1 → 10000 K=2 → 1,000,000,000 K=3 →
10,000,000,000,000 K=100 → 10^400
ͬͱޮྑ͘ܭࢉͰ͖ͳ͍͔ʁ ·ͣK=2ͷ߹Λߟ͑Δ
ߟ 2ͭͷάϥϑ͕ͲͪΒޙखউͪͳΒɺશମͰޙखউͪ 2ͭͷάϥϑ͕(ઌखউͪ + ޙखউͪ)ͳΒɺશମͰઌख উͪ ͩͬͨΒɺ(ઌखউͪ + ઌखউͪ)ͷ߹ʁ ࣮શମͰઌखউͪʹͳΔྫޙखউͪʹͳΔྫ࡞ΕΔ
Ұҙʹఆ·Βͳ͍ ʴ ʴ ʹ ʹ ઌखউͪ ޙखউͪ
Ͳ͏͢Δʁ άϥϑΛʮઌखউͪʯͱʮޙखউͪʯʹ·ͱΊΔ͜ͱͰ ͖ͳ͍ Ͱ2छྨఘΊΔͱͯ͠ɺԿछྨ͔ʹάϥϑΛྨ͢Δ ͜ͱͰ͖ͳ͍ͷ͔ʁ
(άϥϑA + άϥϑX)ͷউഊͱ(άϥϑB + άϥϑX)ͷউഊ ͕ɺXʹ͔͔ΘΒ͍ͣ࣌͠ɺάϥϑA,BΛʮຊ࣭తʹ͠ ͍ʯͱ͢Δ (ޙखউͪ+ޙखউͪ)ͱ(ޙखউͪ+ઌखউͪ)Ұҙʹఆ·Δ ͷͰɺޙखউͪͲ͏͠ʮຊ࣭తʹ͍͠ʯ ͳͷͰɺઌखউͪͲ͏͠ʮຊ࣭తʹ͍͠ʯͷͰʁͱ
Ծఆࣦͯ͠ഊͨ͠ͷ͕ࠓͷঢ়ଶ ઌखউͪͷάϥϑ1छྨʹ·ͱΊΔ͜ͱͰ͖ͳ͍ʂ
ͱ͜ΖͰ ͖ͬࣔͨ͞͠Α͏ʹɺҎԼͷ2ͭͷάϥϑ͘͠ͳ͍ ͳͥʁ
ͱ͜ΖͰ ࠨͷάϥϑޙखউͪʹ͔͠ߦ͚ͳ͍͕ɺӈͷάϥϑޙखউͪͱઌखউͪͷ ྆ํʹߦ͚Δ ޙखউͪͱઌखউ͕ͪ྆ํඋΘΓ࠷ڧʹݟ͑Δ(ӈͷάϥϑͷํ͕ࣗ༝͕ߴͦ ͏)
͡Ό͋ ͜ͷάϥϑߋʹ࠷ڧͳͷͰʁ
͡Ό͋ ͜ͷάϥϑߋʹ࠷ڧͳͷͰʁ
࣮ શDAGͲ͏͠ޓ͍ʹ͘͠ͳ͍͜ͱ͕ࣔͤΔ(TODO) ͳͷͰগͳ͘ͱ༗ݶछྨͰͳ͍
΄΅શDAGʁ શDAGಉ࢜ҟͳΔ͜ͱ͕Θ͔ͬͨ ͡Ό͋શDAGʹ1ݸΛՃ͑ͨͷʁ શDAGͷ͏͍͔ͪͭ͘ͷʹɺ1͔Βล͕ੜ͍͑ͯ ΔɺΑ͏ͳάϥϑ
ྫ͑ ͜Ε1͔Β1, 3ͷDAGʹߦ͚ΔΑ͏ͳάϥϑ
ྫ͑ ͜Ε1͔Β1, 3ͷશDAGʹߦ͚ΔΑ͏ͳάϥϑ ͜Ε2ͷશDAGͱ͍͠
ͳͥʁ 3ͷશDAGʹߦͬͯɺ༗རʹͳΔ͜ͱͳ͍ ͳͥͳΒɺ૬ख࣍ͷखͰ2ͷશDAGҠಈͰ͖ΔͨΊ ݁ہɺ1ͷશDAGߦ͔͘͠ͳ͍ͨΊɺ2ͷશDAGͱຊ࣭తʹ͍͠ ͜ͷลΛͬͯ༗རʹͳΔ͜ͱͳ͍
΄΅શDAG {a, b, c, …}ʹߦ͚Δͱ͢Δͱɺ=ʮ{a, b, c, …}ʹ ग़ͯ͜ͳ͍࠷খͷਖ਼ʯͷશDAGͱຊ࣭తʹ͍͠ ྫ͑{1,
3, 4, 5}ʹߦ͚ΔͳΒ=2 ͳ͔ͥͱ͍͏ͱɺ3,4,5ͷͲΕʹߦ͘ͷɺ࣍ͷखͰ૬खʹ2 ʹ͞ΕΔͷͰɺ༗རʹͳΕͳ͍
Ұൠʹ ઌड़ͷنଇΑΓؼೲతʹɺશͯͷάϥϑશDAGͷͲΕ͔ͱ ຊ࣭తʹ͍͠ =kͷશDAGͱຊ࣭తʹ͍͠άϥϑΛlevel (k-1)ͷάϥ ϑͱݺͿ (0-indexedͰͳ͍ͱ৭ʑ໘) άϥϑͷlevel࠶ؼతʹܭࢉͰ͖ΔͨΊɺO(N+M)ͰٻΊΒΕΔ
݁ہ (aͷશDAG + bͷશDAG)ͱมܗͰ͖Δ ͜Εͷඞউఆ؆୯(a = b͔൱͔) O(N+M)
K=3 K=3ͷ߹ʁ ࠷ॳͷ۪ղΛࢥ͍ग़͢ͱɺෳݸͷάϥϑ1ݸͷάϥϑʹ·ͱΊΒΕΔ ͦͯ͠ɺk=2ͳΒશͯͷάϥϑશDAGมܗͰ͖Δ͜ͱΛ͏ (άϥϑA + άϥϑB + άϥϑC)= {άϥϑA
+ (άϥϑBͱάϥϑCΛ·ͱΊͨͷ)}= (શDAG + άϥϑB + άϥϑC) Αͬͯk=2ͷ࣌ͱಉ͡نଇͰશDAGมܗͯ͠Α͍
K=4 K=4ͷ߹ʁ (άϥϑA + άϥϑB + άϥϑC + άϥϑD)= {άϥϑA+(άϥϑBͱάϥϑCͱάϥϑDΛ·ͱΊͨͷ)}=
(શDAG + άϥϑB + άϥϑC + άϥϑD)
݁ہ (શDAG + શDAG + … + શDAG)ʹมܗͰ͖Δ ͜ͷඞউఆʁ(࣮NimήʔϜͱ͍͏༗໊ͳܗͷήʔϜʹͳ͍ͬͯΔ…) (શDAG
+ શDAG + … + શDAG)= {(શDAGͱશDAGΛ·ͱΊͨͷ) + (K-2ݸͷશDAGΛ·ͱΊͨͷ)}= ʹมܗͰ͖Δ ͱ͜ΖͰɺK=2ͳͷͰ(શDAGͱશDAGΛ·ͱΊͨͷ)શDAGมܗͰ͖Δ ͭ·Γɺ(શDAGͱશDAGΛ·ͱΊͨͷ)ΛશDAGʹมܗͰ͖ΔͱɺK-1ʹؼ ணͰ͖Δ
શDAG + શDAG (શDAGͱશDAGΛ·ͱΊͨͷ)ͱ͍͠શDAGͷ ʁ ࣮ɺ(a+1)ͱ(b+1)Λ·ͱΊΔͱɺ(a xor b)+1 ͱ͘͠ͳΔ(TODOઆ໌) ͭ·Γɺlevel
aͱlevel bΛ߹͢Δͱɺlevel (a xor b)ʹͳΔ (͜Ε͕0-indexedʹͨ͠ཧ༝)
ղ๏ ͱ͍͏͜ͱɺlevel a + b + …ͳΒɺlevel (a xor b
xor …) ͱ͍͠ ͜Ε͕0͔Ͳ͏͔ผ͢Ε͑ O(K(N+M))ʂ