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
Knowledge Distillation for BERT
Search
Scatter Lab Inc.
April 17, 2020
Research
0
2.2k
Knowledge Distillation for BERT
Scatter Lab Inc.
April 17, 2020
Tweet
Share
More Decks by Scatter Lab Inc.
See All by Scatter Lab Inc.
SimCLR: A Simple Framework for Contrastive Learning of Visual Representations
scatterlab
0
3.7k
Adversarial Filters of Dataset Biases
scatterlab
0
2.2k
Sparse, Dense, and Attentional Representations for Text Retrieval
scatterlab
0
2.2k
Weight Poisoning Attacks on Pre-trained Models
scatterlab
0
2.1k
Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval
scatterlab
0
2.4k
Beyond Accuracy: Behavioral Testing of NLP Models with CheckList
scatterlab
0
2.2k
Open-Retrieval Conversational Question Answering
scatterlab
0
2.2k
What Can Neural Networks Reason About?
scatterlab
0
2.2k
Exploring the Limits of Transfer Learning with Unified Text-to-Text Transformer
scatterlab
0
2.1k
Other Decks in Research
See All in Research
メタヒューリスティクスに基づく汎用線形整数計画ソルバーの開発
snowberryfield
3
620
[依頼講演] 適応的実験計画法に基づく効率的無線システム設計
k_sato
0
170
Tiaccoon: コンテナネットワークにおいて複数トランスポート方式で統一的なアクセス制御
hiroyaonoe
0
130
言語処理学会30周年記念事業留学支援交流会@YANS2024:「学生のための短期留学」
a1da4
1
260
Weekly AI Agents News! 9月号 論文のアーカイブ
masatoto
1
140
ミニ四駆AI用制御装置の事例紹介
aks3g
0
180
テキストマイニングことはじめー基本的な考え方からメディアディスコース研究への応用まで
langstat
1
150
LiDARとカメラのセンサーフュージョンによる点群からのノイズ除去
kentaitakura
0
180
言語と数理の交差点:テキストの埋め込みと構造のモデル化 (IBIS 2024 チュートリアル)
yukiar
4
910
PetiteSRE_GenAIEraにおけるインフラのあり方観察
ichichi
0
190
ダイナミックプライシング とその実例
skmr2348
3
480
さんかくのテスト.pdf
sankaku0724
0
510
Featured
See All Featured
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
28
4.3k
The Straight Up "How To Draw Better" Workshop
denniskardys
232
140k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
26
1.8k
Designing for Performance
lara
604
68k
How GitHub (no longer) Works
holman
311
140k
GraphQLとの向き合い方2022年版
quramy
44
13k
Code Reviewing Like a Champion
maltzj
520
39k
Designing for humans not robots
tammielis
250
25k
The Success of Rails: Ensuring Growth for the Next 100 Years
eileencodes
44
6.9k
Documentation Writing (for coders)
carmenintech
66
4.5k
What’s in a name? Adding method to the madness
productmarketing
PRO
22
3.2k
"I'm Feeling Lucky" - Building Great Search Experiences for Today's Users (#IAC19)
danielanewman
226
22k
Transcript
,OPXMFEHF%JTUJMMBUJPOGPS#&35 .BDIJOF-FBSOJOH4PGUXBSF&OHJOFFS
ݾର ݾର 1BUJFOU,OPXMFEHF%JTUJMMBUJPOGPS#&35.PEFM$PNQSFTTJPO %JTUJM#&35 BEJTUJMMFEWFSTJPOPG#&35TNBMMFS GBTUFS DIFBQFSBOEMJHIUFS
'BTU#&35B4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF %ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI
1BUJFOU,OPXMFEHF%JTUJMMBUJPO GPS#&35.PEFM$PNQSFTTJPO 1BUJFOU,OPXMFEHF%JTUJMMBUJPOGPS#&35.PEFM$PNQSFTTJPO
• &./-1"DDFQUFE • .JDSPTPGU%ZOBNJDT"*3FTFBSDI • ୭ୡ#&35୷दب • ӝઓ,OPXMFEHF%JTUJMMBUJPOҗܰѱ1BUJFOUܳਊೣ • 4PGU5BSHFU݅ਊೞঋҊܲۨযীࢲبणೞب۾ೣ
1BUJFOU,OPXMFEHF%JTUJMMBUJPOGPS#&35.PEFM$PNQSFTTJPO 1,%GPS#FSU.PEFM$PNQSFTTJPO
1BUJFOU,OPXMFEHF%JTUJMMBUJPOGPS#&35.PEFM$PNQSFTTJPO णۚ • ࣁоMPTTܳઙೣ • )BSEUBSHFUীೠ$SPTT&OUSPQZ-PTT • 4PGU5BSHFU 5FBDIFSNPEFMPVUQVU ীೠ$SPTT
&OUSPQZ-PTT • рۨযٜ<$-4>షSFQSFTFOBUJPOٜীೠ OPSNBMJ[FE.FBO4RVBSFE-PTT
ण٣పੌ 1BUJFOU,OPXMFEHF%JTUJMMBUJPOGPS#&35.PEFM$PNQSFTTJPO • 4UVEFOU7BSJBUJPO#&35@ #&35@ • #&35@ CBTF ীࢲLѐۨয۽4UVEFOUܳୡӝച •
1,%4LJQFWFSZLMBZFSীࢲण • 1,%-BTUMBTULMBZFSীࢲण
Ѿҗ 1BUJFOU,OPXMFEHF%JTUJMMBUJPOGPS#&35.PEFM$PNQSFTTJPO ؘఠ݆ਸٸੜؽ
Ѿҗ 1BUJFOU,OPXMFEHF%JTUJMMBUJPOGPS#&35.PEFM$PNQSFTTJPO
Ѿҗ 1BUJFOU,OPXMFEHF%JTUJMMBUJPOGPS#&35.PEFM$PNQSFTTJPO
%JTUJM#&35 BEJTUJMMFEWFSTJPOPG#&35 TNBMMFS GBTUFS DIFBQFSBOEMJHIUFS %JTUJM#&35 BEJTUJMMFEWFSTJPOPG#&35TNBMMFS GBTUFS DIFBQFSBOEMJHIUFS
• /FVS*14"DDFQUFE • )VHHJOHGBDF • 4UVEFOUݽ؛ীࢲUPLFOUZQFFNCFEEJOHT QPPMFSMBZFSઁ %JTUJM#&35 BEJTUJMMFEWFSTJPOPG#&35TNBMMFS GBTUFS
DIFBQFSBOEMJHIUFS %JTUJM#&35
%JTUJM#&35 BEJTUJMMFEWFSTJPOPG#&35TNBMMFS GBTUFS DIFBQFSBOEMJHIUFS णۚ • -PTT • 4UVEFOU5FBDIFS$SPTT&OUSPQZ •
.BTLFE-.-PTT • $PTJOF&NCFEEJOH-PTT • 4UVEFOU-BZFSܳпفѐۨযೞաঀஂೣ
Ѿҗ %JTUJM#&35 BEJTUJMMFEWFSTJPOPG#&35TNBMMFS GBTUFS DIFBQFSBOEMJHIUFS
Ѿҗ %JTUJM#&35 BEJTUJMMFEWFSTJPOPG#&35TNBMMFS GBTUFS DIFBQFSBOEMJHIUFS
'BTU#&35 "4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF 'BTU#&35B4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF
• 5FODFOU3IJOP#JSE&MJUF5SBJOJOH1SPHSBNীࢲಎ٬߉ইࢿػ֤ޙ • 4QFFE"DDVSBDZܳઑоמೠ#&35.PEFMઁউ • "EBQUJWF*OGFSFODFооמೠ#&35 4FMG%JTUJMMJOHNFDIBOJTN 'BTU#&35B4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF 'BTU#&35
'BTU#&35B4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF णۚ
'BTU#&35B4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF णۚ • णࣽࢲ • 1SFUSBJOJOH'JOF5VOJOH4FMGEJTUJMMBUJPO • 4FMG%JTUJMMBUJPOزউ୭࢚ਤ#SBODI৬ೞਤ#SBODI$SPTT&OUSPQZ҅ • 6ODFSUBJOUZѐ֛بੑ
• п-BZFS#SBODI߹۽6ODFSUBJOUZ҅ • оࢸ6ODFSUBJOUZоծਸࣻ۾"DDVSBDZо֫
Ѿҗ 'BTU#&35B4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF
Ѿҗ 'BTU#&35B4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF
Ѿҗ 'BTU#&35B4FMGEJTUJMMJOH#&35XJUI"EBQUJWF*OGFSFODF5JNF
%ZOB#&35 %ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI %ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI
• )VBXFJ/PBIsT"SL-BC • XJEUI EFQUIߑೱਵ۽"EBQUJWFೠ#&35ઁউ • 'VMM#&35.PEFMXJEUIBEBQUJWF#&35XJEUIEFQUIBEBQUJWF#&35ള۲ %ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI %ZOB#&35
%ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI णۚ
%ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI णۚ
%ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI णۚ
%ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI णۚ • 8JEUIण • 5FBDIFS4PGU5BSHFUҗ$SPTTFOUSPQZ ߬٬ۨয )JEEFO-BZFSח.4& • %BUBBVHNFOUBUJPOਊ
• %FQUIण • NPE E N@E ੋۨযܳESPQ • FYNVMUJUQMJFSח ਸESPQ • MPTTחXJEUI৬эؘXFJHIUחܴ • ݄݃ীHSPVOEUSVUIMBCFMਸ$SPTT&OUSPQZܳݡৈࢲGJOFUVOJOH
Ѿҗ %ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI
Ѿҗ %ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI
Ѿҗ %ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI
Ѿҗ %ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI
Ѿҗ %ZOB#&35%ZOBNJD#&35XJUI"EBQUJWF8JEUIBOE%FQUI
хࢎפ✌ ୶оޙژחҾӘೠݶઁٚইېোۅ۽োۅࣁਃ .BDIJOF-FBSOJOH4PGUXBSF&OHJOFFS &NBJMVLKBF!TDBUUFSMBCDPLS