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
Federated Learning 連合学習
Search
Sponsored
·
Your Podcast. Everywhere. Effortlessly.
Share. Educate. Inspire. Entertain. You do you. We'll handle the rest.
→
regonn
March 24, 2023
Technology
1.8k
3
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
Federated Learning 連合学習
Regonn @関西Kaggler交流会 2023/03/24
regonn
March 24, 2023
More Decks by regonn
See All by regonn
Allora Network が実現する 「予測の経済圏」
regonn
0
23
Submission with Numerai Compute using GCP Batch
regonn
0
680
HelloMoon x Orca=HelloOrca
regonn
0
180
The Challenge of Automatically Configuring Numerai Compute Environment
regonn
0
140
PLATEAUをデータベースに取り込んでみた
regonn
0
1.8k
古参Numerai参加者の戯言
regonn
1
2.6k
Nim + Cloud Run + TerraformでGCPにNimのサーバーレス環境を作る
regonn
1
700
S+ Camera Basicと似た環境をラズパイから構築してみる
regonn
0
570
衛星画像のゴルフ場写り込み判定AIモデルを作る
regonn
0
870
Other Decks in Technology
See All in Technology
プロンプト_きのこカンファレンス2026_LT
yurufuwahealer
0
130
デジタル・デザイン構想 by Sayaka Ishizuka
y150saya
0
190
AWS Summit 2026で見えたSIerにとっての Amazon Quickの位置づけ
maf_0521
0
160
初めてのDatabricks勉強会
taka_aki
2
230
MySQL & MySQL HeatWave Report - June 2026
freshdaz
0
310
4人目のSREはAgent
tanimuyk
0
380
Mastraエージェント、どのクラウドにデプロイする?
minorun365
PRO
2
150
完全自律ロボットを作りたくて、先に開発を自律させた話(ROS Japan UG #63 LT)
rryz09
0
100
cccccc
moznion
0
1.7k
アラート調査向けAIエージェントの本番導入とその後/AI Agents for Alert Investigation: Production Deployment and After
taddy_919
1
360
#エンジニアBooks 30分でわかる 「技術記事を書く技術」 / engineer-books 2026-06-30
jnchito
1
180
最近評価が難しくなった
maroon8021
0
210
Featured
See All Featured
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
162
16k
Practical Orchestrator
shlominoach
191
11k
[RailsConf 2023] Rails as a piece of cake
palkan
59
6.7k
Game over? The fight for quality and originality in the time of robots
wayneb77
1
220
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
508
140k
"I'm Feeling Lucky" - Building Great Search Experiences for Today's Users (#IAC19)
danielanewman
230
23k
Self-Hosted WebAssembly Runtime for Runtime-Neutral Checkpoint/Restore in Edge–Cloud Continuum
chikuwait
0
640
The Curious Case for Waylosing
cassininazir
1
420
Chrome DevTools: State of the Union 2024 - Debugging React & Beyond
addyosmani
10
1.2k
How to Get Subject Matter Experts Bought In and Actively Contributing to SEO & PR Initiatives.
livdayseo
0
150
The innovator’s Mindset - Leading Through an Era of Exponential Change - McGill University 2025
jdejongh
PRO
1
220
Put a Button on it: Removing Barriers to Going Fast.
kastner
60
4.3k
Transcript
Regonn @ؔKagglerަྲྀձ 2023/03/24 Federated Learning ࿈߹ֶश
Regonn (@regonn_haizine) • Kaggle Competition Expert • Kaggle ؔ࿈ Podcast
(with ΧϨʔͪΌΜ) • ౡࠜͷϑϦʔϥϯε • ϒϩοΫνΣʔϯɾϝλόʔεɾػցֶश(Ӵσʔλɾ)ͷࣄҊ݅ • Solana Japan Orca(DEX: ࢄऔҾॴ) ͷ CM • ڈ͔Βۀ࢝Ίͯ·͢
࠷ۙ AI ͷਐา֮·͍͠ • ChatGPT-4 • Hugging Face ͷ͓͔͛Ͱɺత(Task)ʹ߹ͬͨɺֶशࡁΈϞσϧΛར༻͠ ͘͢ͳ͖͍ͬͯͯΔ
औಘ͕͍͠σʔλᶃ • ΤοδσόΠεσʔλ • ௨৴ଳҬͳͲͷ੍ݶ • શͯͷηϯαʔσʔλऔಘ͍͠
औಘ͕͍͠σʔλᶄ • ҩྍσʔλɾސ٬σʔλ • ϓϥΠόγʔ • ଞࣾؒͰͷڞ༗
Federated Learning(FL) ࢄͯ͠อ࣋͞ΕͨσʔλΛऩूɾू͢Δ͜ͱͳ͘ػցֶशϞσϧͷ܇࿅ʹ ༻͍Δํ๏
Federated Learning(FL)ͷछྨ • ਫฏ࿈߹ֶश • ਨ࿈߹ֶश • ࿈߹సҠֶश
ਫฏ࿈߹ֶश • ͦΕͧΕ͕ɺಉ͡ಛྔͰɺҟͳΔσʔληοτΛ͍࣋ͬͯͨ߹ • ΤοδσόΠεͰ͜ͷλΠϓ
ਫฏ࿈߹ֶश ᶃ ᶄ ᶅ ᶃϩʔΧϧͰֶश ᶄதԝαʔόʔʹଛࣦޯใΛૹΔ ᶅதԝαʔόʔͰΞϯαϯϒϧతॲཧ(ޯฏۉԽ) ᶆߋ৽͞ΕͨϞσϧΛશͯฦ͢ ᶆ
ਨ࿈߹ֶश • ͦΕͧΕಛྔҟͳΔ͕ɺಉ͡λʔήοτͷσʔλΠϯσοΫε͕ଘࡏ͢ Δ߹ ߴ ϩʔϯ ސ٬A xx yy ௨৴ྔ
՝ֹۚ ސ٬A xx yy ΫϨδοτΧʔυͷσϑΥϧτ༧ଌ σϑΥϧτ ސ٬A 1
ਨ࿈߹ֶश ߴ ϩʔϯ ސ٬A xx yy ސ٬C zz ww ௨৴ྔ
՝ֹۚ ސ٬A kk mm ސ٬B aa bb σϑΥϧτ ސ٬A 1 ސ٬D 0 ※ݫີͰͳ͍ தԝαʔόʔͰ҉߸伴Λ࡞Γެ։伴Λɺ ͦΕͧΕͷϩʔΧϧͰ҉߸Խ͠ɺ ΞϊχϚΠζ͞ΕͨσʔλͰֶश ߴ ϩʔϯ ௨৴ྔ աֹۚ σϑΥϧτ ސ٬A xx yy kk mm 1
࿈߹సҠֶश • ಛྔλʔήοτҟͳΔ߹ • ࣅ͍ͯΔυϝΠϯྖҬ • αΠζ͕େ෯ʹҟͳΔ߹
Federated Learningͷใु • ͦΕͧΕͷɺϞσϧߩݙʹԠͯ͡ใुΛܭࢉ͢ΔϩδοΫݚڀ͞Εͯ ͍Δ
Numerai • ۚ༥σʔλͷਨ࿈߹ֶशతͳͷ͕Ͱ͖͍ͯΔ • શੈքגࣜࢢͷגՁͷಈ͖ΛNumeraiଆ͕࡞ͨ͠Target 0~1Ͱ༧ଌ͢Δ • Numerai Signals ͰɺಛྔࣗͰ༻ҙͯ͘͠Δ
• ࣮ࡍʹσʔλͰϑΝϯυ͕ӡ༻͞ΕͯɺརӹͷҰ෦ΛϞσϧͷߩݙʹԠ͡ ͯ͞ΕΔ
Federated Learning ϥΠϒϥϦؔ࿈ • େ͖Ίͷͷͩͱ • https://github.com/tensorflow/federated • TensorflowFederated: Tensorflow
ͷެࣜ Federated ϥΠϒϥϦ • https://github.com/OpenMined/PySyft • PySyft: PyTorch Λ͏߹ͩͬͨΒɺͪ͜ΒΛ͏ • ͓ͦΒ͘ɺϥΠϒϥϦ͚ͩͰͳ͘αʔϏεࠓޙ૿͖͑ͯͦ͏
ࢀߟਤॻ Federated Learning: ϓϥΠόγʔอޢԼʹ͓͚Δػցֶश
ऴΘΓ