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
アットホームな分析基盤の作り方 / Homemade Machine Learning Too...
Search
Yuichiro Someya
July 23, 2018
Programming
1
910
アットホームな分析基盤の作り方 / Homemade Machine Learning Toolkits
分析基盤トーク #1
https://daft.connpass.com/event/93036/
Yuichiro Someya
July 23, 2018
Tweet
Share
More Decks by Yuichiro Someya
See All by Yuichiro Someya
にんげんがさき 基盤はあと / Developers over ML platform
ayemos
0
13k
機械学習をスモールスタートさせる方法 / small machine learning
ayemos
3
1.9k
サービス開発、機械学習、クラウド / the trinity of machine learning
ayemos
0
3.3k
成長を止めない機械学習のやり方 / Don't stop 'til you get enough (data).
ayemos
15
5.1k
AWS で加速する機械学習 / Accelerate Machine Learning with AWS
ayemos
1
300
クックパッドの機械学習基盤 2018 / Machine Learning Platform at Cookpad ~ 2018 ~
ayemos
15
19k
PyTorchとCaffe2とONNXと深層学習モデルのデプロイについて
ayemos
1
2.9k
クックパッドにおけるAWS GPUインスタンスの利用事例 / Powering by AWS GPU Instances in Cookpad Inc
ayemos
0
390
How we use GPUs in Cookpad
ayemos
0
120
Other Decks in Programming
See All in Programming
LangGraphでのHuman-in-the-Loopの実装
os1ma
3
850
大公開!iOS開発の悩みトップ5 〜iOSDC Japan 2024〜
ryunakayama
0
180
BasicBasic認証
sadnessojisan
5
3.1k
iOSの隠されたAPIを解明し、開発効率を向上させる方法/iOSDC24
noppefoxwolf
2
120
From Spring Boot 2 to Spring Boot 3 with Java 22 and Jakarta EE
ivargrimstad
0
1.1k
Modular Monolith Go Server with GraphQL Federation + gRPC
110y
1
570
LangChainでWebサイトの内容取得やGitHubソースコード取得
shukob
0
130
Ebitengineの1vs1ゲーム WebRTCの活用
ponyo877
0
350
Hono・Prisma・AWSでGeoなAPI開発
nokonoko1203
5
630
rails_girls_is_my_gate_to_join_the_ruby_commuinty
maimux2x
0
160
どうしてこうなった?から理解するActive Recordの関連の裏側
willnet
5
530
現代のVueとTypeScript - 型安全の活用術
minako__ph
4
3.1k
Featured
See All Featured
A designer walks into a library…
pauljervisheath
201
24k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
22
1.7k
jQuery: Nuts, Bolts and Bling
dougneiner
61
7.4k
What the flash - Photography Introduction
edds
67
11k
Designing the Hi-DPI Web
ddemaree
278
34k
Making the Leap to Tech Lead
cromwellryan
128
8.8k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
248
20k
StorybookのUI Testing Handbookを読んだ
zakiyama
25
5k
[Rails World 2023 - Day 1 Closing Keynote] - The Magic of Rails
eileencodes
28
1.6k
Raft: Consensus for Rubyists
vanstee
135
6.5k
How to Create Impact in a Changing Tech Landscape [PerfNow 2023]
tammyeverts
42
2k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
123
18k
Transcript
ΞοτϗʔϜͳ ੳج൫ͷ࡞Γํ BZFNPT
ࣗݾհ છ୩༔Ұ<:VJDIJSP4PNFZB> ౦ژۀେֶେֶӃܭࢉֶम࢜ ΫοΫύουגࣜձࣾݚڀ։ൃ෦ ϦαʔνΤϯδχΞ݄d ػցֶशج൫
Ϩγϐσʔλͷੳ UXJUUFSDPN!BZFNPT@Z HJUIVCDPNBZFNPT IUUQTXXXBZFNPTNF
ΫοΫύουݚڀ։ൃ෦ ݄ʹൃ ໊࣌ͷϝϯόʔ ݄ݱࡏࠃʹ໊ ւ֎ʹ໊
ΫοΫύουͱػցֶश .ZϑΥϧμ ͓ؾʹೖΓػೳ ͷϨγϐΛࣗಈΧςΰϦྨ ʮྉཧ͖Ζ͘ʯεϚʔτϑΥϯͷྉཧࣸਅΛΧϨϯμʔܗࣜ ͰৼΓฦΓ ࡐྉ໊ͷਖ਼نԽ
ʑ ৄ͘͠IUUQTUFDIMJGFDPPLQBEDPN
BHFOEB ΫοΫύουͷੳج൫ͷհ ੳج൫ͷ͜Ε͔Β
ج൫հͦͷ;PPFZ
None
ੳͱ࣮ݧΛ֤͕ࣗࣗ༝ʹߦ͏ͨΊͷ࠷ݶͷج൫
ج൫հͦͷ̍;PPFZ ܭࢉػڥߏங༻$IBU#PU ܭࢉػڥͷ࡞ʗ্ཱͪ͛ʗఀࢭʗআ ΞΠυϧঢ়ଶͷܭࢉػΛࣗಈఀࢭ
None
$6%" DV%// $6%" DV%// %FFQ-FBSOJOH".*+BO FUD 0VUHPJOH8FCIPPLT
TMBDL@VJE6 DPNNBOEDSFBUFXPSLCFODI MBNCEB*OWPLF 71$ 4VCOFU /BNFλά ڥಛ༗ͷઃఆ͜͜ͰΔ ΠϯελϯεʹATMBDL@VJEAΛ λά͚ͯ͠ॴ༗ऀΛཧ͢Δ DSFBUFXPSLCFODI ;PPFZ
ATTIBZFNPTXPSLCFODIEOTDPNA ;PPFZ *ODPNJOH8FCIPPLT TTI
;PPFZ 4UPQ ΞΠυϧͳΠϯελϯεΛࣗಈఀࢭ
ج൫հͦͷ̍;PPFZ ͷԾ*NBHF %FFQ-FBSOJOH".* Λར༻Մೳ Πϯελϯε࡞ͷೖΓޱΛҰຊԽ ωοτϫʔΫɺηΩϡϦςΟपΓͷࡉ͔͍ઃఆΛٵऩ
ར༻ଆ*NBHFΛ൪߸ͰબͿ͚ͩ ࣾΠϯϑϥͷมԽʹରͯ͠πʔϧͷΞοϓσʔτͰରԠ ւ֎ͷϝϯόʔར༻ IUUQTBXTBNB[PODPNKQNBDIJOFMFBSOJOHBNJT
ج൫հͦͷ̍;PPFZ ֤͕ࣗSPPUΛऔΕΔΠϯελϯεΛ͍ग़͢ গʑલ࣌తʁ ॊೈੑߴ͘ɺ࣮ݧஈ֊ͷڥͱ͍ͯ͠৺͕Α͍ ࣗಈఀࢭ͋ΔͷͰ֤ࣗ҆৺ͯ͠ඞཁͳ্ཱ͚ͩͪ͛ΒΕΔ
ʮ࣮ݧʯΛؚΊαϙʔτ͢Δڞ௨ج൫Λ࣋ͨͳ͍ͷͰ ʮ࣮ݧʯʮຊ൪ʯͷ1PSUBCJMJUZผ్ิڧ͢Δඞཁ͕͋Δ
ج൫հͦͷσʔλج൫
None
͍ΘΏΔ%8) 42-ϕʔεͷ&5-ج൫
σʔλج൫ ͜ΕओʹผνʔϜ σʔλج൫άϧʔϓ ʹΑΔཧ "NB[PO3FETIJGUʹΑΔ%8)ߏங .Z42-3FETIJGUͷΠϯϙʔτδϣϒΛίʔυཧ
ৄ͘͠IUUQTUFDIMJGFDPPLQBEDPNFOUSZ
σʔλج൫ͷར༻ঢ়گ %8)͕ਆ &5- ࣮ݧͷʹ֤͕ࣗߦ͍ɺεΫϦϓτԽ͢Δࣄ͕ଟ͍ ڞ௨&5-ج൫ͷར༻ɺதؒσʔλͷڞ༗ͳͲʹର͢Δχʔζ ͦ͜·Ͱେ͖͘ͳ͍
ݱঢ়
ج൫հͦͷίʔυཧ
ج൫հͦͷίʔυཧ ج൫ʁ DPPLJFDVUUFSEBUBTDJFODF ࣾGPSL Λར༻ ϓϩδΣΫτߏͷςϯϓϨʔτ
ATSDEBUBA ATSDNPEFMAσʔλੜɺֶशεΫϦϓτ AEBUBA ANPEFMTAσʔλɺϞσϧ HJUʹDIFDLJOͤͣɺ4ͱTZOD͢Δ ANBLFTZOD@EBUB@UP@TA
ج൫հͦͷίʔυཧ IUUQTHJUIVCDPNEPDLFSTDJFODFDPPLJFDVUUFSEPDLFS TDJFODF DPPLJFDVUUFSEBUBTDJFODFΛࢀߟʹͨ͠ςϯϓϨʔτ EPDLFSΛར༻͠ɺϓϩδΣΫτͷ࣮ݱੑΛ͞ΒʹߴΊΔ OPUFCPPLͷ্ཱͪ͛
1PSUGPSXBSEߦ͏UBSHFU
ج൫·ͱΊ ݱঢ়౷ҰԽ͞ΕͨػցֶशϓϥοτϑΥʔϜΛར༻͍ͯ͠ͳ͍ ࣗπʔϧͷΈ߹Θͤ %8)Ͱ͓͓ΉͶճ͍ͬͯΔ ج൫ͦͷͷͷཧίετͳͲߟྀͭͭ͠ਐΊͨ݁Ռ ݱঢ় ൺֱతΧδϡΞϧͳӡ༻ͱͳ͍ͬͯΔ
ੳج൫ͷࠓޙ ʙΞοτϗʔϜͳੳج൫͔Βେ౷Ұج൫ͷಓͷΓʙ ͔͜͜Βߟ
ੳج൫ͷࠜຊత՝ (FOFSJDBOE&YUFOTJCMF
ੳج൫ͷࠜຊత՝ 5'9"5FOTPS'MPX#BTFE1SPEVDUJPO4DBMF.BDIJOF-FBSOJOH1MBUGPSN ΑΓ l5IFNBDIJOFMFBSOJOHQMBUGPSNNVTUCFHFOFSJD FOPVHIUPIBOEMFUIFNPTUDPNNPOTFUPGMFBSOJOH UBTLTBTXFMMBTCFFYUFOTJCMFUPTVQQPSUPOFP⒎ BUZQJDBMVTFDBTFTz ҙ༁ ػցֶशج൫แׅత͔֦ͭுՄೳͰ͋Δ͖
(FOFSJDBOE&YUFOTJWF ʰ࣮ݧ͔Βຊ൪ӡ༻·Ͱɺ౷߹తͳڥͰػցֶशʱ ͠Α͏ͱ͍ͯ͠Δ (FOFSJDͳΔ͘ଟ͘ͷϢʔεέʔεʹ ಁաతʹ ରԠ͍ͨ͠ ྑ͍நԽͱ"1*֦ॆͷؤுΓ͖ͬͱେม
&YUFOTJWFྫ֎తͳέʔεʹରͯ͠ গͳ͍࿑ྗͰ ରԠ͍ͨ͠ ҙͷίʔυΛࠩ͠ࠐΊΔॴΛ༻ҙ͢Δɺͱ͔ʜ
(FOFSJDBOE&YUFOTJWF αϙʔτ͢ΔͷόϦΤʔγϣϯͱͷઓ͍ଓ͘ ྨʗճؼ͘Β͍ͷநԽ͔Β࿙Εͯ͠·͏෦͕ແࢹͰ͖ͳ͍ ྔʹͳ͍ͬͯΔ ཁग़య ج൫ͱ͍͏ܕʹΊΔ͜ͱͰࣦΘΕΔॊೈੑͷ૯ྔΛ༧ଌ͢Δ͜
ͱग़དྷͳ͍ ཁग़య
ߟ ج൫େมͳͣͳͷͰ ࡞Δਓͱ͏ਓ͕૿͑ͯίϛϡχςΟ͕ ͢Δඞཁ͕͋Δ ҰํɺνʔϜͷεέʔϧʹ࠷దԽ͞ΕͨπʔϧΛར༻ʗࣗͯ͠ࡁ·ͤ Δͱ͍͏બࢶ͋Δ ͦͷΑ͏ͳݱ͔Β౷Ұ͞Εͨੳج൫ʹΑͬͯͨΒ͞ΕΔϝϦο
τ͕Πϝʔδ͠ʹ͍͘
ߟ ͷͰɺ લड़ͨ͠Α͏ͳ՝Λ্๊͑ͨͰ ੳج൫͕ଧͪग़͢ϝϦοτԿ͔ɺΛߟ͑Δ ͷ͕େࣄͩͱࢥ͏ ػցֶशΛج൫ͳ͠Ͱεέʔϧͤ͞Δͷ͕ ෆՄೳʹۙ͘ ͍͔͠Βɺ
ಋೖʹ͏ίετ՝Λड͚ೖΕͯͰج൫ʹैͬͯ։ൃ͢Δͱ͍͏બΛ͖͢ɺ ͳͷ͔ εέʔϧ͕͍͠ͱ͍͏࣮ײΛ ۀքશମ͕ ࠓͷ࣌Ͱ ಘ͍ͯΔͷ͔ ڞ௨ج൫Խ͢Δ͜ͱʹΑͬͯಘΒΕΔߴڃͳػೳ "#ςετɺࢹͳͲ ͷັྗΛԡ͠ग़͢ͷ͔ ʑ
ߟ ʮσʔλͷྲྀΕ͕͋ΓɺγεςϜ͕σʔλͱڞʹ͢ΔΑ͏ͳγε ςϜʯ·ͰࢹΛ͛Δͱʜ ػցֶशΛ༻͍ͳ͍ ݕࡧΤϯδϯٞͷର ͱͳΔͷ͕ඞવ ݕࡧΤϯδϯࣗମಠࣗਐԽͯ͠ΔͷͰ৽ͨͳ
ιϦϡʔγϣϯΛଧͪग़͢༨ͳ͍͔͠Εͳ͍͕ʜ ͱ͍͑χϡʔϥϧݕࡧΈ߹ΘͤͨγεςϜʹର͢Δ χʔζࠓޙ૿͍͑ͯͣ͘ ͜͜·Ͱ&YUFOTJCMFʹͳΕΔ͔ ͳΔ͖͔ IUUQTXXXBNB[PODPKQ#VJMEJOH*OUFMMJHFOU4ZTUFNT-FBSOJOH&OHJOFFSJOHFCPPLEQ##82)3
·ͱΊ ΫοΫύουݚڀ։ൃ෦͕ར༻͍ͯ͠Δੳج൫Λհ ෦ʹج൫୲͕͍ͯɺϢʔεέʔεʹরΒ͠߹Θͤͳ͕Βۀվળ ݱঢ়ࣗπʔϧͷΈ߹Θͤ %8)Ͱ͓͓ΉͶຬ ੳج൫ͷ͜Ε͔Β
ݱͱͯͨ͘͠͞ΜͷϢʔεέʔεΛൃ৴͍ͯ͘͠ ػցֶशͷ๊͑Δ՝Λ۩ମԽ͠ɺϝοηʔδͱ͍ͯ͑ͯ͘͠ͷ͕ ॏཁͰେมͦ͏