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
iclr2020deepsemi-supervisedanomalydetectionyama...
Search
Yamato.OKAMOTO
June 14, 2020
Technology
0
140
iclr2020deepsemi-supervisedanomalydetectionyamatookamoto-200531022507.pdf
Yamato.OKAMOTO
June 14, 2020
Tweet
Share
More Decks by Yamato.OKAMOTO
See All by Yamato.OKAMOTO
いまAI組織が求める企画開発エンジニアとは?
roadroller
2
1.5k
Slide ICCV2023 Constructing Image Text Pair Dataset from Books
roadroller
0
110
第11回 全日本コンピュータビジョン勉強会 CVPR2022 "A Self-Supervised Descriptor for Image Copy Detection"
roadroller
0
640
第9回 全日本コンピュータビジョン勉強会 発表資料
roadroller
0
630
第七回全日本コンピュータビジョン勉強会 A Multiplexed Network for End-to-End, Multilingual OCR
roadroller
1
960
部下のマネジメントはAI開発に学べ
roadroller
0
160
Domain Generalization via Model-Agnostic Learning of Semantic Features NeurIPS’19 読み会 in 京都
roadroller
0
280
ICML’2019 読み会in京都 Federated Learningの研究動向
roadroller
0
110
CVPR2019@Long Beach 参加速報(本会議)
roadroller
0
130
Other Decks in Technology
See All in Technology
GCASアップデート(202508-202510)
techniczna
0
340
re:Invent 2025の見どころと便利アイテムをご紹介 / Highlights and Useful Items for re:Invent 2025
yuj1osm
0
690
累計5000万DLサービスの裏側 – LINEマンガのKotlinで挑む大規模 Server-side ETLの最適化
ldf_tech
0
190
CloudComposerによる大規模ETL 「制御と実行の分離」の実践
leveragestech
0
180
新米エンジニアをTech Leadに任命する ー 成長を支える挑戦的な人と組織のマネジメント
naopr
1
360
dbtとAIエージェントを組み合わせて見えたデータ調査の新しい形
10xinc
7
1.8k
現場の壁を乗り越えて、 「計装注入」が拓く オブザーバビリティ / Beyond the Field Barriers: Instrumentation Injection and the Future of Observability
aoto
PRO
1
940
窓口業務を生成AIにおまかせ!Bedrock Agent Coreで実現する自治体AIエージェント!
rayofhopejp
0
160
LLM APIを2年間本番運用して苦労した話
ivry_presentationmaterials
10
7.9k
設計に疎いエンジニアでも始めやすいアーキテクチャドキュメント
phaya72
27
18k
re:Inventに行くまでにやっておきたいこと
nagisa53
0
1k
猫でもわかるAmazon Q Developer CLI 解体新書
kentapapa
1
320
Featured
See All Featured
Cheating the UX When There Is Nothing More to Optimize - PixelPioneers
stephaniewalter
285
14k
Site-Speed That Sticks
csswizardry
13
940
XXLCSS - How to scale CSS and keep your sanity
sugarenia
249
1.3M
Leading Effective Engineering Teams in the AI Era
addyosmani
8
720
Building Applications with DynamoDB
mza
96
6.7k
Reflections from 52 weeks, 52 projects
jeffersonlam
355
21k
VelocityConf: Rendering Performance Case Studies
addyosmani
333
24k
Why Our Code Smells
bkeepers
PRO
340
57k
The Web Performance Landscape in 2024 [PerfNow 2024]
tammyeverts
10
910
Designing Dashboards & Data Visualisations in Web Apps
destraynor
231
54k
What's in a price? How to price your products and services
michaelherold
246
12k
Refactoring Trust on Your Teams (GOTO; Chicago 2020)
rmw
35
3.2k
Transcript
2020/6/14 Yamato OKAMOTO ICLRΦϯϥΠϯಡΈձ Deep Semi-supervised Anomaly Detection
ࣗݾհʢ͘!!ʣ ɹԬຊେʢ͓͔ͱ·ͱʣ • ژେֶඒೱݚڀࣨͰύλʔϯೝࣝΛݚڀͯ͠म࢜՝ఔमྃ • ΦϜϩϯͰ৽نࣄۀΛܦݧޙɺ͍·ࣾձγεςϜࣄۀ෦ͷݚڀॴϦʔμʔ • ເژΛϙετɾγϦίϯόϨʔʹ͢Δ͜ͱɺؔͷίϛϡχςΟΛڧԽ͍ͨ͠ ɹ㱺 ژͷมਓύϫʔΛੈքʹΒ͠Ί͍ͨ
Twitter : RoadRoller_DESU ҆৺҆શͳࣾձͷ࣮ݱʹ͚ͯɺ ࠷ۙ Anomaly Detection ʹڵຯΞϦ
Anomaly Detection ͋Δ͋Δ ఆٛࠔ • ҟৗʹ༷ʑͳόϦΤʔγϣϯ͕͋Δ • ҟৗݕग़͍͚ͨ͠ͲʮWhat is ҟৗʁʯ͕ఆٛͰ͖ͳ͍
ֶशσʔλ͕ೖखࠔ • ҟৗ໓ଟʹൃੜ͠ͳ͍ʢ※ සൟʹൃੜ͢ΔΠϕϯτҟৗ͡Όͳ༷ͯ͘ʣ • ѹతʹҟৗσʔλ͕ෆͯ͠ػցֶश͕ࠔ ैདྷख๏ɿਖ਼ৗΛఆٛ͢Δ • ʮWhat is ҟৗʁʯͷఆٛΛఘΊΔɺҟৗσʔλͷֶशఘΊΔ • ͦͷΘΓʮWhat is ਖ਼ৗʁʯͷఆٛΛֶशͯ͠ɺʮNot ਖ਼ৗʯΛҟৗͱఆ͢Δ
Anomaly Detection ͷैདྷݚڀ Deep One-Class Classification (ICML’18) • ਖ਼ৗσʔλͷΈΛ༻͍ͯɺClassifierͳΓAutoEncoderͳΓΛैདྷ௨Γʹֶश •
͜ͷͱ͖ɺಛྔ͕࣍ݩ෦ۭؒʹऩଋ͢ΔΑ͏LOSSΛՃ͑Δ • ਖ਼ৗσʔλͳΒٿʹ͢ΔͣͳͷͰɺٿ͔Β֎ΕͨσʔλΛҟৗͱఆ͢Δ ୈҰ߲ʹΑͬͯٿʹ͕ԡ͠ࠐ·ΕΔ cɿ ٿͷத৺ʢͨͩ͠≠0ʣ nɿֶश͢Δਖ਼ৗσʔλͷ
Anomaly Detection ͷධՁ؍ ͲΕ͚ͩਖ਼֬ʹҟৗΛݕͰ͖͔ͨʁ • ਖ਼ৗσʔλΛਖ਼ৗͱఆͯ͠ɺҟৗσʔλΛҟৗͱఆ͢Δਫ਼ ԼྲྀλεΫΛअຐ͠ͳ͍͔ʁ • ԼྲྀλεΫ͕͋Δ߹ɺҟৗݕػೳͷՃʹΑͬͯѱӨڹ͕ͳ͍͔Ͳ͏͔ •
ྫ͑ɺ10ΫϥεͷࣈࣝผثʹɺਤܗͳͲࣈҎ֎͕ೖྗ͞Εͨͱ͖ҟৗͱఆ͢Δػ ೳΛ͚Ճ͍͑ͨͤͰɺैདྷͷ10Ϋϥεࣝผੑೳ͕Լ͢ΔͱࠔΔ ad-hoc͔post-hoc͔ʁ • ҟৗݕ͢ΔͨΊʹϞσϧߏֶशํ๏·Ͱม͑Δඞཁ͕͋Δ͔ʁ • ·ͨɺLOSSΛޙ͔Β͚͚̍ͭͩͯ͠Ճֶश͢Δ͚ͩͰOK͔ʁ • ͲͪΒ͕ྑ͍ѱ͍ͳͲҰ֓ʹݴ͑ͳ͍͕ɺpost-hocͷํ͕ѻ͍͍͢ɻ
հจͷ֓ཁ ʮSemi-supervisedʹֶश͠Α͏ʂʯ Anomaly Detection ͷݚڀUnsupervised͕ओྲྀͷΑ͏ͩ Ͱɺֶश༻ͷҟৗσʔλ͕ೖखࠔͩͱͯ͠ɺ ӡ༻Λଓ͚ͯͨΒҟৗσʔλʹ͍ͣΕग़ձ͏ͣ ͳΒɺͦΕΒগྔͷҟৗσʔλΛͬͯɺ Semi-supervisedʹֶशͨ͠ํ͕ྑ͍ͷͰʁ ※Semi-supervisedͷAnomaly
Detectionݚڀඇৗʹগͳ͍
ఏҊख๏ ʮLOSSʹ߲Λ̍ͭՃ͠·ͨ͠ʯ Deep One-Class Classification (ICML’18) ͷLOSSʹSemi-supervisedͷ߲Λ̍ͭՃ • ࣮ಉ͡ஶऀͰͨ͠ɻࣗͷݚڀΛࣗͰΞοϓσʔτͨ͠ܗʹͳΔɻ ͠ҟৗσʔλʹग़ձͬͨΒɺ
ٿͷ֎ଆʹߦ͘Α͏ֶश͢Δ mɿsemi-supervisedʹֶश͢Δσʔλ yj ɿਖ਼ৗorҟৗͷϥϕϧ
࣮ݧ݁Ռ ॎ࣠ɿҟৗσʔλͷݕग़ੑೳ ʢHigher is Betterʣ Unsupervised Semi-supervised ԣ࣠ɿSemi-supervisedͰڭࢣ͖ͷҟৗσʔλΛֶशׂͨ͠߹ ఏҊख๏ MNISTɺFashion-MNISTɺCIFAR-10ͷσʔληοτͰධՁ
• ̍Ϋϥεͱਖ਼ৗͱఆٛͯ͠ɺAutoEncoderʴఏҊख๏ͰಛྔදݱΛֶश • Γͷ̕ΫϥεΛೖྗͨ͠ͱ͖ɺҟৗͱఆͰ͖Δ͔Ͳ͏͔ධՁ ੑೳվળΛ֬ೝ
·ͱΊͱߟ ਂֶशʹΑΔ Semi-supervised ͳ Anomaly Detection ख๏ΛఏҊ • ॳΊͯͰͳ͍ͱࢥ͏͕ɺਂֶशʹΑΔAnomaly DetectionͰsemi-supervised͍͠
• ͔ͨ͠ʹࣾձ࣮Λߟ͑Δͱɺ͜ͷઃఆద • ख๏γϯϓϧͰɺpost-hocͳͷͰѻ͍͍͢ • ࠓճԼྲྀλεΫ͕AE͕ͩͬͨɺClassificationͩͱͲ͏ͳΔ͔ʁ • Anomaly DetectionͷධՁσʔληοτͬͯଞʹͳ͍ͷ͔ͳɺɺɺɺ ʢ͍ͭ·ͰMNISTʹΑΔධՁ͕ଓ͘ͷͩΖ͏͔ʣ