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
機械学習をスモールスタートさせる方法 / small machine learning
Search
Yuichiro Someya
November 06, 2018
Programming
3
2.1k
機械学習をスモールスタートさせる方法 / small machine learning
https://d3m.connpass.com/event/104858/
Yuichiro Someya
November 06, 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
14k
アットホームな分析基盤の作り方 / Homemade Machine Learning Toolkits
ayemos
1
990
サービス開発、機械学習、クラウド / the trinity of machine learning
ayemos
0
3.5k
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
5.2k
AWS で加速する機械学習 / Accelerate Machine Learning with AWS
ayemos
1
330
クックパッドの機械学習基盤 2018 / Machine Learning Platform at Cookpad ~ 2018 ~
ayemos
15
20k
PyTorchとCaffe2とONNXと深層学習モデルのデプロイについて
ayemos
1
3k
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
430
How we use GPUs in Cookpad
ayemos
0
170
Other Decks in Programming
See All in Programming
Things You Thought You Didn’t Need To Care About That Have a Big Impact On Your Job
hollycummins
0
250
Domain-centric? Why Hexagonal, Onion, and Clean Architecture Are Answers to the Wrong Question
olivergierke
3
940
なぜGoのジェネリクスはこの形なのか? - Featherweight Goが明かす設計の核心
qualiarts
0
230
品質ワークショップをやってみた
nealle
0
590
あなたとKaigi on Rails / Kaigi on Rails + You
shimoju
0
170
What's new in Spring Modulith?
olivergierke
1
160
タスクの特性や不確実性に応じた最適な作業スタイルの選択(ペアプロ・モブプロ・ソロプロ)と実践 / Optimal Work Style Selection: Pair, Mob, or Solo Programming.
honyanya
3
190
Developer Joy - The New Paradigm
hollycummins
1
330
はじめてのDSPy - 言語モデルを『プロンプト』ではなく『プログラミング』するための仕組み
masahiro_nishimi
3
6k
オープンソースソフトウェアへの解像度🔬
utam0k
17
3.1k
大規模アプリのDIフレームワーク刷新戦略 ~過去最大規模の並行開発を止めずにアプリ全体に導入するまで~
mot_techtalk
1
470
モテるデスク環境
mozumasu
3
690
Featured
See All Featured
Keith and Marios Guide to Fast Websites
keithpitt
411
23k
Evolution of real-time – Irina Nazarova, EuRuKo, 2024
irinanazarova
9
990
Unsuck your backbone
ammeep
671
58k
Producing Creativity
orderedlist
PRO
347
40k
Agile that works and the tools we love
rasmusluckow
331
21k
BBQ
matthewcrist
89
9.8k
Intergalactic Javascript Robots from Outer Space
tanoku
273
27k
Optimising Largest Contentful Paint
csswizardry
37
3.5k
Templates, Plugins, & Blocks: Oh My! Creating the theme that thinks of everything
marktimemedia
31
2.5k
10 Git Anti Patterns You Should be Aware of
lemiorhan
PRO
657
61k
Reflections from 52 weeks, 52 projects
jeffersonlam
353
21k
Designing Dashboards & Data Visualisations in Web Apps
destraynor
231
53k
Transcript
ػցֶशΛεϞʔϧελʔτ ͤ͞Δํ๏ ΫοΫύουגࣜձࣾછ୩༔Ұ %BUB%SJWFO%FWFMPQFS.FFUVQ
ࣗݾհ છ୩༔Ұ<BZFNPT> ΫοΫύουגࣜձࣾ৽ଔೖࣾ ݚڀ։ൃ෦ΤϯδχΞ ػցֶशج൫ɺը૾ೝࣝܥͷݚڀ։ൃ
ΫοΫύουͱػցֶश ΫοΫύουݚڀ։ൃ෦ ݄ʹൃ ࢲ͕ଐ͞Εͨͷಉ࣌ظ ໊࣌ͷϝϯόʔ
݄ݱࡏࠃʹ໊ ւ֎ʹ໊ ʰສͷϨγϐσʔλΛ׆༻͠ɺϢʔβʔʹՁΛಧ͚Δʱ
ࠓͷτϐοΫ ػցֶश ओʹਂֶश Λ εϞʔϧελʔτͤ͞Δํ๏ ͳͥεϞʔϧελʔτ͕ඞཁ͔ Ͳ͏Δͷ͔
ࠓͷτϐοΫ ٕज़ʗέʔεελσΟগͳΊɺ ίϯηϓτଟΊͷʹͳΓ·͢ աڈʹٕͨ͠ज़తͳͪ͜Β IUUQTTQFBLFSEFDLDPNBZFNPT
ਂֶशͱεϞʔϧελʔτ
ਂֶशͱεϞʔϧελʔτ 4NBMMTUBSU εϞʔϧελʔτ 4UBSUTNBMM
lUIJOLJOHCJH TUBSUJOHTNBMM BOETDBMJOHGBTUz IUUQTKJNDBSSPMMDPNJOOPWBUJPOUIJOLCJHTUBSUTNBMMTDBMFGBTU
5IJOLCJHJEFOUJGZUIFMPOHUFSNUSBOTGPSNBUJWFUSFOET JODMVEJOHTJHOJpDBOUJOEVTUSZDIBOHF CVTJOFTTNPEFM EJTSVQUJPO ྫ͑ɺʮσΟʔϓϥʔχϯάΛ͍ͬͯ͘ʯͱܾΊΔ
4UBSUTNBMM1JDLBOVNCFSPGTNBMM FYQFSJFOUJBM PSJFOUBUFEQSPKFDUTUPCFHJO5IJTXJMMHJWFZPVCFUUFS EFQUIPGJOTJHIU ͍ͬͯͨ͘Ίʹখ͍͞ϓϩδΣΫτΛ͜ͳ͠ɺֶͿ
4DBMFGBTU%FUFSNJOFXIJDIBSFBTOFFEUPCFUBDLMFE pSTUJOUFSNTPGNPWJOHGPSXBSE%FWFMPQUIFBCJMJUZUP UBLFZPVSbQSPUPUZQJOH`PGTLJMMTFOIBODFNFOUGSPNUIF TNBMMTDBMFQSPKFDUTJOUPGVMMqFEHFEPQFSBUJPOT ༗ͳϓϩδΣΫτΛબͼɺຊ֨తʹՔಇͤ͞Δ
5IJOLCJH ྫ͑ ʮਂֶशΛ͍ͬͯ͘ʯͱܾΊΔ 4UBSUTNBMMখ͍͞ϓϩδΣΫτ ࣮ݧ Λ͜ͳ͠ɺֶͿ 4DBMFGBTU༗ͳϓϩδΣΫτΛຊ֨తʹՔಇͤ͞Δ
ਂֶशͱ4UBSUTNBMM 4UBSUTNBMMখ͍͞ϓϩδΣΫτΛ͜ͳ͠ɺֶͿ ਂֶशͱ4UBSUTNBMMͷ૬ੑ͕͍͍ͱࢥΘͤΔ ৽ٕज़Ͱ͋ΓɺԿ͕ͲΕ͘Β͍Ͱ͖Δͷ͔प͞Ε͍ͯͳ͍ ෦తࢼߦࡨޡ͕ඞཁ
લྫͳ͍ͷͰɺޭ ྫϢʔβʔͷՁʹܨ͕Δ ͢Δͷ͔͔Βͳ͍ ֎෦తʏ ٕज़ελοΫͱͯ͠ݟͯૣख़Ͱ͋Δ ӡ༻ͷٕज़తशख़͕ඞཁ
ਂֶशͱ4UBSUTNBMM ҰํͰɺ4UBSUTNBMM͕Γʹͦ͘͏ͳҰ໘͋Δ େྔͷσʔλͱܭࢉث͕ඞཁ ӡ༻ίετ͕ߴ͍ ࠾༻େมͦ͏
IUUQTBJHPPHMFSFTFBSDIQVCTQVC
ਂֶशͱ4UBSUTNBMM ҰํͰɺ4UBSUTNBMM͕Γʹͦ͘͏ͳҰ໘͋Δ େྔͷσʔλͱܭࢉث͕ඞཁ ӡ༻ίετ͕ߴ͍ ࠾༻େมͦ͏
IUUQTBJHPPHMFSFTFBSDIQVCTQVC .-0QTͰ ͕ ղܾͰ͖ͦ͏ͳ
ͱ͜ΖͰ.-0QTͱ ڪΒ͘%FW0QTಉ༷͕ͩ ·ͩఆٛෆ໌ྎ ֶशΞϧΰϦζϜҎ֎ʁ %FW0QTͷ.-൛ʁ ػցֶश͕ɺιϑτΣΞ։ൃࣄۀʹ͓͍ͯՁΛੜΈग़͢͜ͱͷͰ͖Δ
ٕज़ελοΫͰ͋Γଓ͚ΔͨΊʹඞཁͳٕज़ ʰӡ༻ίετͷݮʱ͚͕ͩతͰͳ͍ ྫਂֶशΛεϞʔϧελʔτͤ͞Δҝͷ.-0QT
͜͜·Ͱ·ͱΊ ਂֶशΛεϞʔϧελʔτ͍ͤͨ͞ ͦͷͨΊʹ.-0QTͷϓϥΫςΟεΛ׆͔ͤͦ͏
εϞʔϧελʔτͷͨΊͷ.-0QT ΫοΫύουͷ߹
ܭࢉثڥ ݄ͷ ݚڀ։ൃ෦һਓ དྷि͔ΒΠϯλʔϯΛਓड͚ೖΕ "84ຊ൪ΞΧϯτʹྑͷ(16Πϯελϯε HYMBSHF ͕
৽ن࡞ͷʹίϛϡχέʔγϣϯ͕ൃੜ ݚڀ։ൃ༻ΞΧϯτΛൃߦ͠ɺӡ༻ͷੵۃతͳԽ
ܭࢉثڥ ݚڀ։ൃ༻ΞΧϯτͷΠϯϑϥΛίʔυཧ ຊ൪ΞΧϯτͷϓϥΫςΟε %FW0QT ʹ฿͏ $IBU#PUΛ௨ͯ͠(16ΠϯελϯεΛ্ཱͪ͛ΔΈΛ࣮
Πϯελϯεͷ࡞ɺٳΠϯελϯεͷࣗಈఀࢭΛඋ ࣮࣭ (16Πϯελϯεݐͯ์Λ࣮ݱ ͍ͭͰ࣮ݧεϞʔϧελʔτ
ܭࢉثڥ ݚڀ։ൃ༻ΞΧϯτͷΠϯϑϥΛίʔυཧ ຊ൪ΞΧϯτͷϓϥΫςΟε %FW0QT ʹ฿͏ $IBU#PUΛ௨ͯ͠(16ΠϯελϯεΛ্ཱͪ͛ΔΈΛ࣮
Πϯελϯεͷ࡞ɺٳΠϯελϯεͷࣗಈఀࢭΛඋ ࣮࣭ (16Πϯελϯεݐͯ์Λ࣮ݱ ͍ͭͰ࣮ݧεϞʔϧελʔτ .-0QT
ػցֶशج൫ͱ4UBSUTNBMM ػցֶशج൫·ͨ৽͍͠ྖҬ4UBSUTNBMM͕ඞཁ ඞཁͳͷج൫ٕज़ͷεϞʔϧͳࢼߦࡨޡ ྫ͑ج൫୲Λ3%ʹஔͯ͠ΈΔ LVCFqPXͳͲͷϓϥοτϑΥʔϜ৫γνϡΤʔγϣϯ ͱͷ૬ੑ͕͋Δ
ՄೳͰ͋Ε ੵۃతʹࢼ͢ɺͬͯΈΔͱΑͦ͞͏
༨ஊʙݕࡧγεςϜʹֶͿʙ ʮσʔλͷྲྀΕ͕͋ΓɺγεςϜ͕σʔλͱڞʹ͢ΔΑ͏ͳγ εςϜʯͱ͍͑ʁ ݕࡧγεςϜͷӡ༻ϓϥΫςΟε͔Β ֶͿ͜ͱଟͦ͏ ΠϯσοΫεͷߏஙɺࣙॻσʔλͷཧ IUUQTXXXBNB[PODPKQ#VJMEJOH*OUFMMJHFOU4ZTUFNT-FBSOJOH&OHJOFFSJOHFCPPLEQ##82)3
·ͱΊ ਂֶशεϞʔϧελʔτ͍ͤͨ͞ ৽͍ٕ͠ज़ͷՄೳੑ ͱ੍ Λ࡞Γͳ͕ΒֶͿ εϞʔϧελʔτ .-0QTͷదͳར༻εϞʔϧελʔτΛॿ͚Δ