Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
RG WIP 1
Search
omochi
February 03, 2017
Research
0
75
RG WIP 1
This is my very first WIP presentation in RG.
(2017.02.03)
omochi
February 03, 2017
Tweet
Share
More Decks by omochi
See All by omochi
RG WIP 2017_08_02
0xomochi
0
87
ISC paper reading 2
0xomochi
0
66
ISC paper reading 1
0xomochi
0
110
Other Decks in Research
See All in Research
生成的情報検索時代におけるAI利用と認知バイアス
trycycle
PRO
0
290
第66回コンピュータビジョン勉強会@関東 Epona: Autoregressive Diffusion World Model for Autonomous Driving
kentosasaki
0
340
生成AIとうまく付き合うためのプロンプトエンジニアリング
yuri_ohashi
0
130
世界モデルにおける分布外データ対応の方法論
koukyo1994
7
1.4k
離散凸解析に基づく予測付き離散最適化手法 (IBIS '25)
taihei_oki
PRO
1
680
超高速データサイエンス
matsui_528
2
380
競合や要望に流されない─B2B SaaSでミニマム要件を決めるリアルな取り組み / Don't be swayed by competitors or requests - A real effort to determine minimum requirements for B2B SaaS
kaminashi
0
720
SREはサイバネティクスの夢をみるか? / Do SREs Dream of Cybernetics?
yuukit
3
380
ブレグマン距離最小化に基づくリース表現量推定:バイアス除去学習の統一理論
masakat0
0
140
Aurora Serverless からAurora Serverless v2への課題と知見を論文から読み解く/Understanding the challenges and insights of moving from Aurora Serverless to Aurora Serverless v2 from a paper
bootjp
6
1.5k
第二言語習得研究における 明示的・暗示的知識の再検討:この分類は何に役に立つか,何に役に立たないか
tam07pb915
0
1.1k
svc-hook: hooking system calls on ARM64 by binary rewriting
retrage
1
100
Featured
See All Featured
Unsuck your backbone
ammeep
671
58k
Build your cross-platform service in a week with App Engine
jlugia
234
18k
Ecommerce SEO: The Keys for Success Now & Beyond - #SERPConf2024
aleyda
1
1.8k
VelocityConf: Rendering Performance Case Studies
addyosmani
333
24k
Paper Plane (Part 1)
katiecoart
PRO
0
4.2k
B2B Lead Gen: Tactics, Traps & Triumph
marketingsoph
0
55
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
26
3.3k
Designing Experiences People Love
moore
144
24k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
133
19k
The Straight Up "How To Draw Better" Workshop
denniskardys
239
140k
Deep Space Network (abreviated)
tonyrice
0
49
Marketing to machines
jonoalderson
1
4.6k
Transcript
ϚϧΣΞྨʹ͚ͨ ੩తղੳʹΑΔಛநग़ *4$#ZVNF HPNBDIBO 3(8*1࠷ऴൃද
ݚڀ֓ཁ • ϚϧΣΞྨͷਫ਼͕มಈ͢Δཁૉ • ಛྔ • ಛநग़ͷํ๏ • ػցֶशΞϧΰϦζϜ •
ΞϧΰϦζϜ͝ͱʹͲͷΑ͏ͳҧ͍͕ग़Δͷ͔ʁ • ੩తղੳͷΈΛߦ͏ 2 ϚϧΣΞྨͷͨΊͷಛநग़෦ͷ։ൃ
3 എܠ
ϚϧΣΞͷݱঢ় 4 [ະϚϧΣΞͷݕମ] [ϚϧΣΞײછͷݕ] ʮຽ࿈ܞϓϩδΣΫτACTIVE ϚϧΣΞ࠷৽ϨϙʔτʯURL<http://www.active.go.jp/active/data/>ΑΓ
ղܾ͢Δ͖՝ • ײછʹؾ͕͍͔ͭͯΒඃ͕֦େ͠ͳ͍Α͏ʹରॲ͢Δ • ײછʹؾ͕͔ͭͣʹରԠ͕ΕΔ͜ͱ͋Δʢաڈʹେྔͷݸਓใ ͕ྲྀग़ͨ͠ࣄྫ͋Γʣ • ϚϧΣΞʹײછ͠ͳ͍ͨΊͷ༗ޮͳରࡦऔΒΕ͍ͯͳ͍ ɹɹɹʮײછ͔ͯ͠ΒʯରॲΛ͢Δͱ͍͏ํ ɹɹɹˠϚϧΣΞʹײછ͢Δલʹݕ͍ͨ͠
5 ϚϧΣΞײછ͕໌͔ͯ͠Βରॲ
ղܾख๏ ػցֶशΛಋೖ͢Δͱʜ • ະͷϚϧΣΞΛࣝผ͢Δ͜ͱ͕Ͱ͖ΔΑ͏ʹͳΔ ੩తղੳͱಈతղੳͷํΛ༻͍Δͱʜ • ղੳΛ্͛Δ͜ͱͰ৽छͷϚϧΣΞʹਝʹରԠ Ͱ͖ΔΑ͏ʹͳΔʢຊݚڀ੩తղੳͷΈʣ 6 ϚϧΣΞղੳʹػցֶशΛ༻͍Δ
7 ػցֶशͱ
ػցֶशͱ 8 ίϯϐϡʔλʹେྔͷσʔλΛֶशͤ͞ ύλʔϯΛݟ͚ͭग़͢͜ͱ αϯϓϧσʔλ ಛநग़ ಛྔ ղੳ ֤छػցֶश ΞϧΰϦζϜ
ʻਤɿػցֶशͷྲྀΕʼ ҰఆҎ্ͷσʔλ͕ඞཁ σʔλ͕࣋ͭಛΛද͢ϕΫτϧ ͲͷΑ͏ͳنଇ͕͋Δ͔ΛௐΔ
ػցֶशͷԠ༻ 9 Ԡ༻ ྨ ŞŽţŦžƃŞƄ ճؼ ࣍ݩݮ ॲཧ༰ • ༩͑ΒΕͨ
σʔλʹϥ ϕϧΛ͚ͭ ͯྨ • ͷྨࣅੑ Λݩʹάϧ ʔϓ͚ • աڈͷσʔ λΛݩʹক དྷͷΛ ༧ଌ • σʔλͷಛ Λҡ࣋͠ ͭͭ࣍ݩΛ Լ͛Δ ׆༻ྫ • ໎ϝʔϧ ͷྨ • खॻ͖จࣈ ͷೝࣝ • ࠂͷ͓͢ ͢Ίػೳ • גՁมಈ • ൢച༧ଌ • ܭࢉͷߴ Խ • ϝϞϦઅ
ػցֶशͷྨ 10 ྨ ڭࢣ͋Γֶश ڭࢣͳֶ͠श ڧԽֶश ֶशํ๏ (༩͑ΒΕΔ ͷ) •
σʔλ • ਖ਼ղ • σʔλ • ෆશͳ͑ ֶश݁Ռ • ະͷσʔλ ʹରͯ͠༧ଌ Λߦ͏ • ະͷσʔλ ͔ΒنଇੑΛ ൃݟ͢Δ • σʔλ͔Β࠷ దղΛൃݟ͢ Δ
ػցֶशͷΞϧΰϦζϜ 11 • TDJLJUMFBSOͰαϙʔτ͞Ε͍ͯΔΞϧΰϦζϜҎԼͷछ ʢTDJLJUMFBSOʹຊݚڀͰ༻͍Δػցֶश༻ϑϨʔϜϫʔΫͷҰछʣ • ྨ $MBTTJGJDBUJPO •
ΫϥελϦϯά $MVTUFSJOH • ճؼ 3FHSFTTJPO • ࣍ݩݮ %JNFOTJPOBMJUZ3FEVDUJPO ʢࢀߟɿTDJLJUMFBSOBMHPSJUINDIFBUTIFFUʣ
12 Ҿ༻ݩʮhttp://scikit-learn.org/stable/tutorial/machine_learning_map/ʯ ΫϥελϦϯά
13 ղੳͷྲྀΕ
ղੳͷྲྀΕ 14 ϚϧΣΞݕମΛऔಘ • Լهͷ63-Ϧετ͔ΒϚϧΣΞݕମΛऔಘ • ʮ.BMXBSF%PNBJO-JTUʯXXXNBMXBSFEPNBJOMJTUDPN • ʮ797BVMUʯIUUQWYWBVMUTJSJVS[OFU63-@-JTUQIQ •
ʮ.BMDEFʯIUUQNBMDEFDPNSTT • ݕମΛμϯϩʔυ͢Δࡍʹ5IF0OJPO3PVUFS τʔΞ Λ༻͍ͨ ɹˠଓܦ࿏Λಗ໊Խ͢Δ͜ͱͰɺϚϧΣΞղੳΛ͍ͯ͠Δ͜ͱΛѱҙͷ͋Δ ૬खʹೝ͞Εͣʹղੳ͕Ͱ͖Δ
ղੳͷྲྀΕʢ՝ʣ 15 ՝ɿσʔλྔΛ૿͢ ɹɹɾݱࡏͷݕମɿʢཧݕମɿ ʣ ɹɹɹˠσʔλྔ͕গͳ͍ͱ͑ΔΞϧΰϦζϜ͕ݶΒΕΔ ɹɹɹɹʢݱஈ֊ͰछͷΞϧΰϦζϜ͕͑Δʣ ՝ɿಉҰϑΝΠϧͷॏෳμϯϩʔυΛ͙ ɹɾμϯϩʔυ͞ΕΔݕମͭͷαΠτʹ͋ΔϦετͷத͔Β ɹɹϥϯμϜʹબΕͨͷͰ͋ΔͨΊɺಉҰϑΝΠϧ͕μϯϩʔυ
ɹɹ͞ΕΔ͜ͱ͕͋Δ
16 μϯϩʔυ͞Εͨ ϚϧΣΞݕମ ^
ղੳͷྲྀΕ 17 όΠφϦσʔλͷٯΞηϯϒϧ • ٯΞηϯϒϥͷҰछͰ͋ΔDBQTUPOFΛ༻͍ͯٯΞηϯϒϧ • ͡ΊPCKEVNQίϚϯυͰٯΞηϯϒϧͨ͠ͷΛ͓͏ͱ ࢥ͍ͬͯͨ ɹˠಛநग़ͷஈ֊Ͱͭ·͍ͮͨͷͰมߋͨ͠ •
DBQTUPOFΛ༻͍ͯٯΞηϯϒϧΛ͢ΔͨΊͷίʔυΛ1ZUIPO Ͱॻ͍͍ͯΔ్த
۩ମྫ(ϚϧΣΞݕମ”0695674e66201ac6de38fab1eba1230c” ͷٯΞηϯϒϧίʔυϦετͷҰ෦) 18 ΞυϨε όΠφϦσʔλ Φϖίʔυ χʔϞχοΫ
۩ମྫ(ϚϧΣΞݕମ”0695674e66201ac6de38fab1eba1230c” ͷٯΞηϯϒϧίʔυϦετͷҰ෦) 19 ΞυϨε όΠφϦσʔλ Φϖίʔυ χʔϞχοΫ 16ਐͷͰ࣮ߦ͞ΕΔίʔυͷ ϓϩηεϝϞϦ্ͷ൪Λද͢
۩ମྫ(ϚϧΣΞݕମ”0695674e66201ac6de38fab1eba1230c” ͷٯΞηϯϒϧίʔυϦετͷҰ෦) 20 ΞυϨε όΠφϦσʔλ Φϖίʔυ χʔϞχοΫ 16ਐͷͰද͞ΕΔσʔλͰ ϓϩηεϝϞϦ্ʹ͋Δ࣮ߦίʔυͷ࣮ଶ
۩ମྫ(ϚϧΣΞݕମ”0695674e66201ac6de38fab1eba1230c” ͷٯΞηϯϒϧίʔυϦετͷҰ෦) 21 ΞυϨε όΠφϦσʔλ Φϖίʔυ χʔϞχοΫ ػցޠͷ༰Λਓ͕ؒཧղ͍͢͠Α͏ ʹΘ͔Γ͘͢༁ͨ͠ӳࣈจࣈྻ
۩ମྫ(ϚϧΣΞݕମ”0695674e66201ac6de38fab1eba1230c” ͷٯΞηϯϒϧίʔυϦετͷҰ෦) 22 ΞυϨε όΠφϦσʔλ Φϖίʔυ χʔϞχοΫ χʔϞχοΫͷॳΊʹ͋ΔӳࣈจࣈྻͰ CPU͕ߦ͏ॲཧͷछྨΛࢦ͢ call
= APIͷݺͼग़͠
ղੳͷྲྀΕ 23 ಛྔΛநग़͢Δ "1*ͷݺͼग़͠ճΛಛྔʹ͠Α͏ͱ͕ͨ͠ʜ • جຊతʹ"1*ݺͼग़͠ϧʔϓͷதͰߦ͏ͨΊ੩తղੳΛ͢Δࡍʹ DBMMͷճΛಛྔͱͯ͋͠·Γҙຯ͕ͳ͍ • ΠϯϙʔτηΫγϣϯʹॻ͔ΕͨใͱরΒ͠߹ΘͤΔඞཁ͋Γ ɹɹΠϯϙʔτηΫγϣϯΛύʔεͯ͠ಘΒΕΔ
ɹɹɹɹɹɹ"1*ϦετΛಛྔͱ͢Δ
ղੳͷྲྀΕ • ΫϥελϦϯάͷͨΊͷΞϧΰϦζϜΛ༻͍Δ༧ఆ • .FBO4IJGU ฏۉมҐ๏ • ࢀߟ IUUQTDJLJUMFBSOPSHTUBCMFNPEVMFT
DMVTUFSJOHIUNMNFBOTIJGU ɹɾ7#(.. • ࢀߟ IUUQTDJLJUMFBSOPSHTUBCMFNPEVMFT NJYUVSFIUNMWCHNNDMBTTJGJFSWBSJBUJPOBMHBVTTJBONJYUVSFT 24 ෳͷػցֶशΞϧΰϦζϜͰղੳ
25 ༻͍ΔػցֶशΞϧΰϦζϜ
खॱ 26 ࣮த • طଘͷ63-Ϧετ͔ΒϚϧΣΞݕମΛऔಘ • ϚϧΣΞݕମͷόΠφϦσʔλΛDBQTUPOF Λ༻͍ͯٯΞηϯϒϧ • ಛநग़ͷҝͷϓϩάϥϜ࣮
ະ࣮ • ػցֶश༻ϑϨʔϜϫʔΫͷҰछͰ͋Δ TDJLJUMFBSOΛ༻͍ͯෳͷػցֶश ΞϧΰϦζϜͰϚϧΣΞΛղੳ ະ࣮
ࠓޙͷ༧ఆ • ϚϧΣΞݕମͷσʔλΛDBQTUPOFΛ༻͍ͯٯΞηϯϒϧ • ಛநग़ͷҝͷϓϩάϥϜΛ࣮ • TDJLJUMFBSOΛ༻͍ͯෳͷػցֶशΞϧΰϦζϜͰϚϧΣΞ Λղੳ 27 ࣮͔ΒධՁ·Ͱ
ࠓޙͷܭը 28 2݄ 3݄ 4݄ 5݄ 6݄ Πϕϯτ WIP(RGൃද) WIPதؒ
ಛநग़ ಛྔͷಡΈࠐΈ ݁Ռूܭ ධՁ ৽ςʔϚ
དྷظҎ߱ͷඪ 29 • ࠓֶظطଘͷݚڀΛͳͧͬͨͷ • དྷظҎ߱ΑΓಠࣗੑͷ͋ΔςʔϚΛઃఆ͢Δ • ίʔσΟϯάྗΛ͚ͭΔ • ػցֶश
ϚϧΣΞʹؔ͢ΔࣝΛਂΊΔ • ΑΓଟ͘ͷؔ࿈ݚڀ ࢿྉʹ͋ͨΔ
ࢀߟจݙ • ஶऀɿΫδϥඈߦص ʮ1ZUIPOʹΑΔεΫϨΠϐϯάػցֶशʯग़൛ࣾɿιγϜ ग़൛ɿ • ʮίʔυͷಛʹجͮ͘ѱੑϓϩάϥϜͷྨʯүҪརએʢʣ • ʮޮతͳղੳΛతͱͨࣗ͠ಈϚϧΣΞྨʹؔ͢Δݚڀʯؠଜʢʣ
• ʮ0OUIF4FDVSJUZPG.BDIJOF-FBSOJOHJO.BMXBSF$$%FUFDUJPO"4VSWFZʯ+PTFQI (BSEJOFS 4IJTIJS/BHBSBKB -BODBTUFS6OJWFSTJUZʢʣ • ʮ"CZTT8BUDIFS.BMXBSF%PXOMPBEFSʯ63-IUUQTHJUIVCDPNOUEEL"CZTT 8BUDIFS • ʮ"$5*7&ʢϚϧΣΞରࡦࢧԉʣ)1ʯ63-IUUQXXXBDUJWFHPKQTFDVSJUZNBMXBSF 30
ࠓޙͷ༧ఆ • ϚϧΣΞݕମͷσʔλΛDBQTUPOFΛ༻͍ͯٯΞηϯϒϧ • ಛநग़ͷҝͷϓϩάϥϜΛ࣮ • TDJLJUMFBSOΛ༻͍ͯෳͷػցֶशΞϧΰϦζϜͰϚϧΣΞ Λղੳ 31 ࣮͔ΒධՁ·Ͱ
͝ਗ਼ௌ͋Γ͕ͱ͏͍͟͝·ͨ͠