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
Adallocation-9
Search
Shinya Akiba
August 31, 2016
Research
1
57
Adallocation-9
Shinya Akiba
August 31, 2016
Tweet
Share
More Decks by Shinya Akiba
See All by Shinya Akiba
RS for HP Opt
shinya_akiba
0
3.3k
Mining_Data_Stream
shinya_akiba
0
380
社内勉強会 NIPS2015
shinya_akiba
1
1.1k
Other Decks in Research
See All in Research
湯村研究室の紹介2024 / yumulab2024
yumulab
0
370
Weekly AI Agents News! 10月号 論文のアーカイブ
masatoto
1
460
CUNY DHI_Lightning Talks_2024
digitalfellow
0
270
Weekly AI Agents News! 12月号 論文のアーカイブ
masatoto
0
120
研究を支える拡張性の高い ワークフローツールの提案 / Proposal of highly expandable workflow tools to support research
linyows
0
250
チュートリアル:Mamba, Vision Mamba (Vim)
hf149
6
2k
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
170
Geospecific View Generation - Geometry-Context Aware High-resolution Ground View Inference from Satellite Views
satai
2
150
Composed image retrieval for remote sensing
satai
2
150
20241115都市交通決起集会 趣旨説明・熊本事例紹介
trafficbrain
0
870
VisFocus: Prompt-Guided Vision Encoders for OCR-Free Dense Document Understanding
sansan_randd
1
420
CVPR2024 参加報告
kwchrk
0
150
Featured
See All Featured
The World Runs on Bad Software
bkeepers
PRO
66
11k
The Cult of Friendly URLs
andyhume
78
6.1k
The Psychology of Web Performance [Beyond Tellerrand 2023]
tammyeverts
45
2.3k
Easily Structure & Communicate Ideas using Wireframe
afnizarnur
192
16k
"I'm Feeling Lucky" - Building Great Search Experiences for Today's Users (#IAC19)
danielanewman
226
22k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
251
21k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
27
1.5k
jQuery: Nuts, Bolts and Bling
dougneiner
62
7.6k
CSS Pre-Processors: Stylus, Less & Sass
bermonpainter
356
29k
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
132
33k
Code Reviewing Like a Champion
maltzj
521
39k
Designing for Performance
lara
604
68k
Transcript
0OMJOF.BUDIJOHBOE"E"MMPDBUJPO
"6OJpFE7JFX#JE4DBMJOH"MHPSJUINT 0CKFDUJWF'VODUJPOT6TFEJO1SBDUJDF 1VCMJTIFE3FTVMUTGSPNUIF*OEVTUSZ "EBQUJWF#JETDBMJOH)FVSJTUJDT 5ISPUUMJOH
"EBQUJWF#JETDBMJOH)FVSJTUJDT ͜Ε·ͰɺछྨͷCJETDBMJOHΞϧΰϦζϜΛհͨ͠ɻ "EWFSTBSJBM ΫΤϦʹΛԾఆ͠ͳ͍ ࠷ѱέʔεͷධՁ %JTUSJCVUJPOCBTFE
4UBUJDͳΛԾఆ ࠷ॳ ͘͠લͳͲ ͷαϯϓϦϯάʹΑΔ͕શମͷΛಉ͡ ͱ͍͏Ծఆ ࣮ࡍɺͷதͰಉ͡Ͱͳ͍͠ɺมԽ͍ͯ͘͠ɻ
"EBQUJWF#JETDBMJOH)FVSJTUJDT ఆظతʹϦαϯϓϦϯάSFDPNQVUF-1`TEVBMWBSJBCMFT "EWFSTBSJBMNPEFMΛར༻͢Δɻ< > )ZCSJEBMHPSJUINTΛ͏ɻ<> "UJNFWBSZJOHEJTUSJCVUJPOJOQVUNPEFM<
> )FVSJTUJDͳख๏ GFFECBDLMPPQ TUBUJDEJTUSJCVUJPO OPOTUBUJDEJTUSJCVUJPO
<>'BTU"MHPSJUINTGPS'JOEJOH.BUDIJOHTJO-PQTJEFE #JQBSUJUF(SBQITXJUI"QQMJDBUJPOTUP%JTQMBZ"ET ࠂओͷ༧ࢉ͕ҟͳΔέʔεͰɺΑΓ#BMBODFEͳΞϧΰϦζϜɻ <>.477ͱ0⒐JOF-1ͷIZCSJE "MMPDBUJOHPOMJOFBEWFSUJTFNFOUTQBDFXJUIVOSFBMFTUJNBUFT .BIEJBO :BIPP3FTFBSDI .477Ͱར༻͍ͯͨ͠ЇFGͱ͍͏ؔΛ ЇFЋ G
ͱ͢ΔɻGɺ ͓ͬͨۚ༧ࢉ <>"E$FMM"E"MMPDBUJPOJO$FMMVMBS/FUXPSLT "E$FMMBOFXNPCJMFBEWFSUJTJOHDPODFQU ࣌ؒଳҐஔใʹΑͬͯΧελϚʔΛධՁ
<>"E8PSETBOE(FOFSBMJ[FE0OMJOF.BUDIJOH ڝ߹ൺFͱͳΔΑ͏ͳΞϧΰϦζϜ ҎԼͷ࣮༻తͳ݅ΛԾఆ͠ɺFΛ࣮ݱͨ͠ɻ ೖࡳऀ ࠂओ ɺVTFS͕ࠂΛΫϦοΫͨ͠ͱ͖ͷΈDPTUΛ ࢧ͏ ࠂओɺ͝ͱʹҟͳΔ༧ࢉΛ࣋ͭ
J൪ͷೖࡳऀɺJ ൪ͷೖࡳऀͷೖࡳֹΛࢧ͏ ͭͷΫΤϦ ྫݕࡧΩʔϫʔυ ʹର͠ɺෳͷࠂ͕දࣔ͞ ΕΔ ࠂओ͝ͱͰɺೖࡳճ͕ҟͳΔ
<>3FBM5JNF#JEEJOH"MHPSJUINTGPS1FSGPSNBODF #BTFE%JTQMBZ"E"MMPDBUJPO OPOTUBUJDEJTUSJCVUJPOʹରͯ͠ɺ༗ޮͳख๏ɻ NBSLFUͷมԽʹରͯ͠ɺ ࣌ؒత ཻͷࡉ͔͍CJEΛઃఆɻ ϦΞϧλΠϜͰิਖ਼߲Λߋ৽͍͖ͯ͠ɺ͍࣌ؒͰͷͷ มԽʹରԠɻ
5ISPUUMJOH CJETDBMJOHͱ5ISPUUMJOHͷҧ͍ 5ISPUUMJOHͷత 5ISPUUMJOHͷߟ͑ํͭ "OFXJOQVUNPEFM
0OMJOF,OBQTBDL1SPCMFNXJUIFTUJNBUFT 0QUJNJ[JOHGPSBMMBEWFSUJTFSTTJNVMUBOFPVTMZ #FTUF⒎PSUSFHSFUGSFFUISPUUMJOH
5ISPUUMJOH CJETDBMJOHͱ5ISPUUMJOHͷҧ͍ 6 7 6 7 TDBMF͞ΕͨCJET͔Β ࠷େͷͷΛબͿ ΦʔΫγϣϯʹࢀՃ͢Δ͔
Ͳ͏͔ΛܾΊ͔ͯΒબͿ PSʹTDBMF͢Δ CJETDBMJOH UISPUUMJOH
5ISPUUMJOH CJETDBMJOHͱ5ISPUUMJOHͷҧ͍ 6 7 6 7 TDBMF͞ΕͨCJET͔Β ࠷େͷͷΛબͿ ΦʔΫγϣϯʹࢀՃ͢Δ͔
Ͳ͏͔ΛܾΊ͔ͯΒબͿ PSʹTDBMF͢Δ 5ISPUUMF͞Εͨ ࠂΩϟϯϖʔϯ CJETDBMJOH UISPUUMJOH
5ISPUUMJOH 5ISPUUMJOHͷత ࠂΩϟϯϖʔϯͷ༧ࢉҎ্ʹೖࡳ͠ͳ͍ͨΊ ΩϟϯϖʔϯظؒதɺTNPPUIʹඅফ͍ͤͯͨ͘͞Ί 5ISPUUMJOHͷํ๏
*NQSFTTJPOWʹରͯ͠ɺVΛUISPUUMF͢Δ͔Ͳ͏͔ͷ֬QVΛ V͕Ωϟϯϖʔϯऴྃͱಉ࣌ʹඅফ͖͠ΕΔΑ͏ʹબͿɻ <>Ͱɺ࠷େԽͨ͠తؔΛ࠷େԽ͢ΔΑ͏ͳ0QUJNJ[FE UISPUUMJOHͱ͍͏ߟ͑ํΛಋೖ͍ͯ͠Δɻ
5ISPUUMJOH "OFXJOQVUNPEFM BVDUJPOUJNFQFSBEWFSUJTFSNFUSJDЖΛఆٛ͢Δɻ ྫ͑ɺࠂΩϟϯϖʔϯVʹର͢Δ*NQSFTTJPOWͷ$53 4V Y ЖV W
≥YͷΑ͏ͳYʹ͍ͭͯɺΫΤϦWʹର͢Δظඅফɻ 4V͕ࣄલʹ͔͍ͬͯΔͱ͍͏Ծఆɻ
5ISPUUMJOH "OFXJOQVUNPEFM ༧ࢉ#Vͱ͖ɺЖ $53 ≥НͷΫΤϦʹೖࡳ͢Δɻ ЖΛ$1$ͱ͓͍ͨ߹ɺЖ $1$ ≤НͷΫΤϦʹೖࡳ͢Δɻ
5ISPUUMJOH 0OMJOF,OBQTBDL1SPCMFNXJUIFTUJNBUFT ЗVVͷWBMVFQFSDMJDL ΫϦοΫͨΓͷظച্ 30*Λ࠷େԽ͢ΔΑ͏ͳ,OBQTBDL P⒐JOF Λߟ͑Δɻ 30*Λ࠷େԽ͢Δ͜ͱɺЗV$1$Λ࠷େԽ͢Δ͜ͱͱಉ͡ɻ ЗVࠂΩϟϯϖʔϯ͝ͱʹݻఆͳͷͰɺ$1$Λখ͘͢͞Δɻ
ඪͷඅফΛୡͭͭ͠ɺ30*Λ࠷େԽ͢Δʹɺ$1$Нͷ ΫΤϦWΛUISPUUMF͢Δɻ
5ISPUUMJOH 0OMJOFͰͷΞϧΰϦζϜɻ 0OMJOF,OBQTBDL1SPCMFNXJUIFTUJNBUFT
5ISPUUMJOH 0QUJNJ[JOHGPSBMMBEWFSUJTFSTTJNVMUBOFPVTMZ ͜Ε·Ͱͭͷࠂओ Ωϟϯϖʔϯ ʹ͍ͭͯɺUISPUUMJOHͷ Λߟ͖͑ͯͨɻ࣮ࡍʹɺෳͷࠂओ͕ΦʔΫγϣϯʹࢀ Ճ͍ͯ͠ΔͷͰɺͦΕͧΕͷඅফЖͷಠཱͰͳ͍ɻ ෳࠂओͷࢀՃʹΑΓʜ
ܝࡌॱҐ͕Լ͕Γɺ$53͕མͪΔ TFDPOEQSJDFΦʔΫγϣϯʹΑΓɺ$1$͕࣍ͷॱҐͷVͷ ͷͱͳΔ <>ͰɺଞͷࠂओΛߟྀͨ͠UISPUUMJOHΛఏҊ͍ͯ͠Δ
5ISPUUMJOH #FTUF⒎PSUSFHSFUGSFFUISPUUMJOH ෳࠂओͰͷ࠷దԽΛߟ͑ͨͱ͖ʹɺෆެฏ͞ͷ͕ੜ͡Δɻ ྫ͑ɺ HMPCBMͳ ΫϦοΫ࠷େԽͷͱ͖ɺ༧ࢉͷେ͖ͳࠂओ ͷΫϦοΫ͕૿͑Δɻ <>ͰCFTUF⒎PSUɺ<>ͰSFHSFUGSFFͱ͍͏ख๏Ͱɺ ࠂओ͝ͱʹෆެฏ͕ੜ͡ͳ͍Α͏ʹ͍ͯ͠Δɻ