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
Stanford Covid Vaccine 2nd place solution
Search
Kazuki Fujikawa
June 16, 2021
Science
0
89
Stanford Covid Vaccine 2nd place solution
Stanford Covid Vaccine 2nd place solution
Kazuki Fujikawa
June 16, 2021
Tweet
Share
More Decks by Kazuki Fujikawa
See All by Kazuki Fujikawa
GUI操作LLMの最新動向: UI-TARSと関連論文紹介
kfujikawa
0
1.7k
Enhancing News Recommendation with Transformers and Ensemble Learning
kfujikawa
0
120
H&M Personalized Fashion Recommendation
kfujikawa
0
350
コミュニティ検出 動向紹介
kfujikawa
1
580
Stable Diffusion - Image to Prompts
kfujikawa
4
2.4k
BMS Molecular Translation 3rd place solution
kfujikawa
0
96
ACL2020 best papers
kfujikawa
0
84
Kaggle参加報告: Champs Predicting Molecular Properties
kfujikawa
0
110
NLP@ICLR2019
kfujikawa
0
56
Other Decks in Science
See All in Science
Kaggle: NeurIPS - Open Polymer Prediction 2025 コンペ 反省会
calpis10000
0
310
データベース08: 実体関連モデルとは?
trycycle
PRO
0
1k
俺たちは本当に分かり合えるのか? ~ PdMとスクラムチームの “ずれ” を科学する
bonotake
1
350
Lean4による汎化誤差評価の形式化
milano0017
1
400
データマイニング - グラフデータと経路
trycycle
PRO
1
260
【RSJ2025】PAMIQ Core: リアルタイム継続学習のための⾮同期推論・学習フレームワーク
gesonanko
0
560
データから見る勝敗の法則 / The principle of victory discovered by science (open lecture in NSSU)
konakalab
1
260
風の力で振れ幅が大きくなる振り子!? 〜タコマナローズ橋はなぜ落ちたのか〜
syotasasaki593876
1
180
検索と推論タスクに関する論文の紹介
ynakano
1
110
データベース03: 関係データモデル
trycycle
PRO
1
330
[Paper Introduction] From Bytes to Ideas:Language Modeling with Autoregressive U-Nets
haruumiomoto
0
180
データマイニング - ノードの中心性
trycycle
PRO
0
320
Featured
See All Featured
The Cost Of JavaScript in 2023
addyosmani
55
9.4k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
48
9.8k
The Hidden Cost of Media on the Web [PixelPalooza 2025]
tammyeverts
2
130
Winning Ecommerce Organic Search in an AI Era - #searchnstuff2025
aleyda
0
1.8k
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
141
34k
Discover your Explorer Soul
emna__ayadi
2
1k
The Web Performance Landscape in 2024 [PerfNow 2024]
tammyeverts
12
990
Redefining SEO in the New Era of Traffic Generation
szymonslowik
1
180
Facilitating Awesome Meetings
lara
57
6.7k
Become a Pro
speakerdeck
PRO
31
5.8k
Navigating Team Friction
lara
191
16k
Context Engineering - Making Every Token Count
addyosmani
9
570
Transcript
%4᳒ᰆ ,B[VLJ'VKJLBXB Ꮠუ័୷ΰᏐუ័.PCJMJUZ5FDIOPMPHJFT ,BHHMF$PNQFUJUJPO 4UBOGPSE$PWJE7BDDJOF OEQMBDFTPMVUJPO 5FBN,B[VLJ ,B[VLJ4RVBSFE ,B[VLJ0OPEFSB
,B[VLJ'VKJLBXB
े 4UBOGPSE$PWJE7BDDJOF े ୣணஊ୶୧ணᑁᮉ े உ୩ஙண े 4PMVUJPO "(&/%"
े 4UBOGPSE$PWJE7BDDJOF े ୣணஊ୶୧ணᑁᮉ े உ୩ஙண े 4PMVUJPO "(&/%"
े $07*%ଠடୟୱணಬ᭦ଚଉଘN3/"டୟୱண૾ᔞଇାଘଽ े ༾ᑗଜ᜔ᅴଝജ൙ᛟᧄଜለᬾଙો᷻ᜲᴌႼഋᅌᡡଝჵୄሿଖ े ࿖ᅌଜଛଝག૿ଜᯠẴୄሿଓોରᯔଇାఔ౺૾ᘏ े ࿖ᅌடୟୱண૾ᛛᇡଇାଘ૽౦ങଙඇୄᜲኅଋଽରଙଠᴝ᠊ଠ ଛଅ૽ଙો൏ᮛଇାଘ፡ඇᅌ૾ཉୁାଘଉର े
ଛଠ3/"ᑑᴍ૾൏ᮛଇାଶଋଠ૽ોରញᮌ૾ဍଜ ୣணஊ୶୧ணᑁᮉᦴጳ IUUQTXXXLBHHMFDPNDTUBOGPSEDPWJEWBDDJOFPWFSWJFX 3/"൏ྶଠค༐ଙଠ൏ᮛᴌႼୄቯ࿖ଙ૿ଽఒᕺஒ୷ୄ౬଼ો டୟୱண᷻ᜲଝᄎᠼଘ
े N3/"൏ྶଠค༐ଝଋଽౄ௦ଠᎇ௦ଙଠ෧ᄷᅌୄఒᕺଋଽ े SFBDUJWJUZค༐ଠ෧ᄷᅌ े
[email protected]
@Q)Q)ଙஎୠ୧ୖஐୄตଲᖚྜྷଙଠค༐ଠ෧ᄷᅌ े
[email protected]
@$ݽଙஎୠ୧ୖஐୄตଲᖚྜྷଙଠค༐ଠ෧ᄷᅌ ୣணஊ୶୧ணᑁᮉ୯୩ୟ KWWSVZZZNDJJOHFRPFVWDQIRUGFRYLGYDFFLQHGLVFXVVLRQ
ሷᇶ␒ྕ 6HTXHQFH * * $ $ $ $ * 6WUXFWXUH 3UHGLFWHG/RRS7\SH ( ( ( ( ( 6 6 ሷᇶ␒ྕ UHDFWLYLW\ GHJB0JBS+ GHJB0JB&
े சୟ୩ े .$3.4&ῠஙஐᓍଠ3.4&Ⴅ໎ῡ े ᮨᤚ୷୯୶୩୷୯ῠ1VCMJD-#ῡ े ጃଝ࿚ỿᣨ෪ᄠᕩଠഋ᷶༐ଠN3/"൏ྶ े ଉો4/ᓏ૾ဌଇଁોಘằᅌଠౢ୷୯ଡᯀಃ૽Ḩ༻ଇାଽ
े 4/ᓏଡᮨᤚ୷୯ଠ௨ૺାଽ े ୶୩୷୯ῠ1SJWBUF-#ῡ े ୣணஊ፫᷾௴ଝ௱ᬻଉଘ࿚ỿଇାഋ᷶༐ଠN3/"൏ྶ े 1VCMJD-#ฉᑗો4/ᓏ૾ဌଇ୷୯ଡ፞ᣡᯀಃ૽Ḩ༻ଇାଽ े ߓଅଠᵄ൏ଙ,BHHMFᴛຄଠ1SJWBUF-#ᮣᡴଝஏ୩૾଼ો -#ଠജᮣᡴ૾ᬻୁାଽଅଚଝ ୣணஊ୶୧ணᑁᮉᯀಃ
े 4UBOGPSE$PWJE7BDDJOF े ୣணஊ୶୧ணᑁᮉ े உ୩ஙண े 4PMVUJPO े ,BHHMFୣணஊଝૼଃଽ࿚ỿᡷᚫ
"(&/%"
े 4FRVFODF 4USVDUVSF 1SFEJDUF-PPQ5ZQFୄዥྐྵൕଠଝሄો 3//ῠ-45.(36ῡଙஒ୷சணୠ े IUUQTXXXLBHHMFDPNYIMVMVPQFOWBDDJOFTJNQMFHSVNPEFM உ୩ஙண-45.(36 VWUXFWXUH
VHTXHQFH **$$$$*&8« (PEHGGLQJ (PEHGGLQJ (PEHGGLQJ * * $ SUHGLFWHGBORRSBW\SH (((((66666« ( ( ( /670 *58 UHDFWLYLW\ GHJBS+ GHJB0JBS+ GHJB& GHJB0JB&
े 4USVDUVSF #11ῠ#BTF1BJSJOH1SPCBCJMJUZ.BUSJYῡୄକଘ ୠஙஅୄᑑᇡ े IUUQTXXXLBHHMFDPNNSLNBLSDPWJEBFQSFUSBJOHOOBUUODOO உ୩ஙண(// VHTXHQFH **$$$$*&8« 2+(
2+( 'LVWDQFH0DWUL[ SUHGLFWHGBORRSBW\SH (((((66666« ESS * * $ ( ( ( VWUXFWXUH DGMXVWPHQW *11 UHDFWLYLW\ GHJBS+ GHJB0JBS+ GHJB& GHJB0JB&
े 4UBOGPSE$PWJE7BDDJOF े ୣணஊ୶୧ணᑁᮉ े உ୩ஙண े 4PMVUJPO "(&/%"
4PMVUJPOᑁᮉ IUUQTXXXLBHHMFDPNDTUBOGPSEDPWJEWBDDJOFEJTDVTTJPO
4PMVUJPOᑁᮉ IUUQTXXXLBHHMFDPNDTUBOGPSEDPWJEWBDDJOFEJTDVTTJPO
े ᭲ዝଠ3/"ᑑᴍఒᕺச୪ஐଙଜଽ#11ୄ෪ᄠ े ᮨᤚ୷୯ଠᓝ༗ଉ55"ଝᛟ े ஒ୷ങଙ᭲ዝଠ#11ୄ൙ᛟ 4PMVUJPO%BUB"VHNFOUBUJPO OMP>OPM@
MDBDI<G D@II<½ JIOM<AJG? PK<>F I<NJAO O@MI<AJG?
4PMVUJPOᑁᮉ IUUQTXXXLBHHMFDPNDTUBOGPSEDPWJEWBDDJOFEJTDVTTJPO
े ᭲ዝଠ#11ୄୱஓዷฎଝᵒᥔ 4PMVUJPO-45.(36ῠ,'ῡ VWUXFWXUH VHTXHQFH **$$$$*&8« (PEHGGLQJ
(PEHGGLQJ (PEHGGLQJ * * $ &RQY' 0D[SRRO SUHGLFWHGBORRSBW\SH (((((66666« ( ( ( ESSV /670 *58 UHDFWLYLW\ GHJBS+ GHJB0JBS+ GHJB& GHJB0JB&
े ᭲ዝଠ#11ଙୠஙஅୄᑑᢀ 4PMVUJPO-45.(36ῠ0OPEFSBῡ VHTXHQFH **$$$$*&8« 2+( 2+( 'LVWDQFH0DWUL[ SUHGLFWHGBORRSBW\SH (((((66666«
ESSV * * $ ( ( ( VWUXFWXUH DGMXVWPHQW *11 UHDFWLYLW\ GHJBS+ GHJB0JBS+ GHJB& GHJB0JB&
4PMVUJPOᑁᮉ IUUQTXXXLBHHMFDPNDTUBOGPSEDPWJEWBDDJOFEJTDVTTJPO
े 3//(//ଠ᭲ዝஒ୷4UBDLJOHଠ00'ଙ1TFVEP-BCFMୄ࿚ዹ े ଅାୄଇଝ4UBDLJOHଋଽଚ00'ଝᴝ࿁ଉ1TFVEP-BCFMஒ୷ଠ JNQPSUBODF૾ག૿ଁଜ଼ଋଽଳો୪ୄૺଘ4UBDLJOH 4PMVUJPO1TFVEP-BCFM QSUDQGRPQRUPDO
4PMVUJPO4UBDLJOH H[S;;;LV$( *11
4PMVUJPO4UBDLJOH
4PMVUJPO4UBDLJOH
4PMVUJPOᑁᮉ IUUQTXXXLBHHMFDPNDTUBOGPSEDPWJEWBDDJOFEJTDVTTJPO
े 1VCMJD1SJWBUFଠག૿ଜೖฎၼᣃൕ᷶ῠWTῡ े $7ଙẻଠೖฎၼୄ౬଼ોག૿ଜᅌᧄඁට૾ᘏ૽឴ᯔ 4PMVUJPO-#୧ஏகஜ୧ண 7UDLQ 3ULYDWHWHVW VHTBOHQJWK VHTBVFRUHG VHTBOHQJWK
VHTBVFRUHG 9DOLG VHTBOHQJWK VHTBVFRUHG 7UDLQ VHTBOHQJWK VHTBVFRUHG 6LPXODWH *11 *58 7UDLQ 9DOLG
े $71VCMJD-#ଡჵଁḂଉଘ $7WT1VCMJD-#
े $71SJWBUF-#ଡஒ୷ଝକଘଡဍଉᅸଠྺ े 4UBDLJOHჵଉ $7WT1SJWBUF-#
े ᭲ዝଠ3/"ᑑᴍఒᕺச୪ஐୄଅଚଙોஒ୷ଠᢱႼୄ ག૿ଁዋଋଽଅଚ૾ଙ૿ े 1TFVEP-BCFM 4UBDLJOH૾ṻ႖ଝ፡ඇକ े ῠჟୱஐଝḢଌῡ1VCMJD1SJWBUFଠ୩ୣၼଡག૿૽କ ῠ1VCMJDYߓ1SJWBUFYῡ े
ଅଠ୩ୣၼଡોසᣍଜ୷୯ᄃუଠᴠଝଽᛣᎋଙଡଜଇଏ ῠ-#୧ஏகஜ୧ணଙଡଜଽೖฎଠᣨୄ៍ଉଘῡ े ༷ዝଶ3/"ᑑᴍଠ୯ணଝག૿ଜᴠ૾କᧄᅌῷ 4PMVUJPOରଚଳ