[B! æ©Ÿæ¢°å¦ç¿’] n_knuuã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

n_knuu id:n_knuu

æ©Ÿæ¢°å¦ç¿’ã«é–¢ã™ã‚‹n_knuuã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (102)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

freee vs ãƒžãƒãƒ¼ãƒ•ã‚©ãƒ¯ãƒ¼ãƒ‰ã®ç‰¹è¨±æ¨©ä¾µå®³è¨´è¨Ÿç¬¬ï¼‘å¯©åˆ¤æ±ºã¯AIã®è¦³ç‚¹ã‹ã‚‰ã‚‚èˆˆå‘³æ·±ã„ | STORIAæ³•å¾‹äº‹å‹™æ‰€
æ ªå¼ä¼šç¤¾ãƒžãƒãƒ¼ãƒ•ã‚©ãƒ¯ãƒ¼ãƒ‰ï¼ˆä»¥ä¸‹ã€ŒMFç¤¾ï¼‰ï¼‰ã®ã€ŒMFã‚¯ãƒ©ã‚¦ãƒ‰ä¼šè¨ˆã€ãŒè‡ªç¤¾ã®ç‰¹è¨±ã‚’ä¾µå®³ã—ã¦ã„ã‚‹ã¨ã—ã¦ã€ãƒ•ãƒªãƒ¼æ ªå¼ä¼šç¤¾ï¼ˆä»¥ä¸‹ã€Œfreeeã€ï¼‰ãŒ2016å¹´10æœˆã«æ±äº¬åœ°æ–¹è£åˆ¤æ‰€ã«æèµ·ã—ãŸç‰¹è¨±æ¨©ä¾µå®³è¨´è¨Ÿã§ã™ãŒã€ç¬¬ï¼‘å¯©åˆ¤æ±ºãŒ2017å¹´7æœˆ27æ—¥ã«å‡ºã¾ã—ãŸã€‚ çµæžœã¯ç‰¹è¨±æ¨©è€…ã§ã‚ã‚‹freeeã®æ•—è¨´ã§ã™ã€‚ ã€å‚è€ƒã€‘ è‡ªå‹•ä»•è¨³ã§ç‰¹è¨±ä¾µå®³ãªã—ã€ãƒžãƒãƒ¼ãƒ•ã‚©ãƒ¯ãƒ¼ãƒ‰ãŒfreeeã«å‹è¨´ ä¼šè¨ˆã®ä»•è¨³é …ç›®ã‚’æ©Ÿæ¢°çš„ã«åˆ¤å®šã™ã‚‹ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã‚’å·¡ã£ã¦freeeã¨ãƒžãƒãƒ¼ãƒ•ã‚©ãƒ¯ãƒ¼ãƒ‰ã®é–“ã§äº‰ã‚ã‚ŒãŸç‰¹è¨±è¨´è¨ŸãŒ9ã‚«æœˆã¨ã„ã†æ¯”è¼ƒçš„çŸæœŸã§çµ‚çµã—ãŸã€‚æ±äº¬åœ°è£ãŒå…ˆã»ã©ï¼ˆ7æœˆ27æ—¥åˆå¾Œã«ï¼‰å‡ºã—ãŸåˆ¤æ±ºã¯ã€freeeå´ã®è«‹æ±‚ã‚’æ£„å´ã™ã‚‹ã‚‚ã®ã§ã€ç‰¹è¨±ä¾µå®³ã¯èªã‚ã‚‰ã‚Œãªã„ã¨ã„ã†çµè«–ã ã£ãŸã€‚ ã“ã®è£åˆ¤ã¯ã€ãƒ™ãƒ³ãƒãƒ£ãƒ¼åŒå£«ã®ç‰¹è¨±æ¨©ä¾µå®³è¨´è¨Ÿã¨ã„ã†ã“ã¨ã§ã‚‚æ³¨ç›®ã‚’é›†ã‚ãŸã®ã§ã™ãŒã€AIï¼ˆæ£ç¢ºã«ã¯æ©Ÿæ¢°å¦ç¿’ï¼‰ã‚’åˆ©ç”¨ã—ãŸæŠ€è¡“ã¨ç‰¹è¨±ä¾µå®³ã¨ã„ã†é¢ã‹ã‚‰ã‚‚ã€ãªã‹ãªã‹é¢ç™½ã„åˆ¤æ±ºã§ã™ã€‚ ã¡ãªã¿
n_knuu 2022/01/14
æ©Ÿæ¢°å¦ç¿’

ç‰¹è¨±
ãƒªãƒ³ã‚¯
Accumulated Local Effectsï¼ˆALEï¼‰ã§æ©Ÿæ¢°å¦ç¿’ãƒ¢ãƒ‡ãƒ«ã‚’è§£é‡ˆã™ã‚‹ - Dropout
ã¯ã˜ã‚ã« Partial Dependence ç‰¹å¾´é‡ãŒç‹¬ç«‹ã®å ´åˆ æ•°å¼ã«ã‚ˆã‚‹ç¢ºèª PDã®å®Ÿè£… ç‰¹å¾´é‡ãŒç›¸é–¢ã™ã‚‹å ´åˆ PDãŒã†ã¾ãæ©Ÿèƒ½ã—ãªã„åŽŸå› Marginal Plot Marginal Plotã®æ•°å¼ Marginal Plotã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ Maginal Plotã®å®Ÿè£… Accumulated Local Effects ALEã®ã‚¢ã‚¤ãƒ‡ã‚¢ ALEã¯ã†ã¾ãæ©Ÿèƒ½ã™ã‚‹ã®ã‹ ALEã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ALEã®å®Ÿè£… ALEã®æ•°å¼ ã¾ã¨ã‚ Appendixï¼šç·šå½¢å›žå¸°ãƒ¢ãƒ‡ãƒ«ã®å ´åˆ å‚è€ƒæ–‡çŒ® ã“ã®è¨˜äº‹ã‚’ãƒ™ãƒ¼ã‚¹ã«ã—ãŸç™ºè¡¨è³‡æ–™ã§ã™ï¼ speakerdeck.com ã¯ã˜ã‚ã« Random Forestã‚„ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°ãªã©ã®ãƒ–ãƒ©ãƒƒã‚¯ãƒœãƒƒã‚¯ã‚¹ãƒ¢ãƒ‡ãƒ«ã¯ã€äºˆæ¸¬æ€§èƒ½ãŒé«˜ã„ä¸€æ–¹ã§è§£é‡ˆæ€§ãŒä½Žã„ã¨ã„ã†ãƒˆãƒ¬ãƒ¼ãƒ‰ã‚ªãƒ•ã‚’æŠ±ãˆã¦ã„ã¾ã™ã€‚ ã“ã‚Œã‚’å…‹æœã™ã‚‹ãŸã‚ã«ã€ãƒ–ãƒ©ãƒƒã‚¯ãƒœãƒƒã‚¯ã‚¹ãƒ¢ãƒ‡ãƒ«ã«å¾Œã‹ã‚‰è§£é‡ˆæ€§ã‚’ä¸Žãˆã‚‹ã€Œæ©Ÿæ¢°å¦ç¿’ã®è§£é‡ˆæ‰‹æ³•
n_knuu 2021/09/13
æ©Ÿæ¢°å¦ç¿’

èª¬æ˜Žå¯èƒ½æ€§
ãƒªãƒ³ã‚¯
Rustã§æ‰±ãˆã‚‹æ©Ÿæ¢°å¦ç¿’é–¢é€£ã®ã‚¯ãƒ¬ãƒ¼ãƒˆ2021 - Stimulator
- ã¯ã˜ã‚ã« - æœ¬è¨˜äº‹ã§ã¯ã€Rustã§æ‰±ãˆã‚‹æ©Ÿæ¢°å¦ç¿’é–¢é€£ã‚¯ãƒ¬ãƒ¼ãƒˆã‚’ã¾ã¨ã‚ã‚‹ã€‚ æ™®æ®µPythonã§æ©Ÿæ¢°å¦ç¿’ãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆã‚’é‚è¡Œã™ã‚‹äººãŒRustã«ç§»è¡Œã™ã‚‹äº‹ã‚’æƒ³å®šã—ã¦æ›¸ããƒ¡ãƒ¢æ›¸ãã«ãªã‚‹ãŒã€ã‚‚ã—ã‹ã™ã‚‹ã¨é•·ã‚‰ãRustã§MLã‚’ã‚„ã£ã¦ã„ãŸäººã¨è¦–ç‚¹ã®é•ã„ãŒã‚ã‚‹ã‹ã‚‚ã—ã‚Œãªã„ã€‚ è¿½è¨˜ï¼š2021/02/24 repositoryã«ã—ã¾ã—ãŸã€‚ã“ã¡ã‚‰ã‚’éšæ™‚æ›´æ–°ã—ã¾ã™ github.com è¿½è¨˜ï¼›2021/07/26 GitHub Pagesã§ã‚¦ã‚§ãƒ–ã‚µã‚¤ãƒˆã«ã—ã¾ã—ãŸ vaaaaanquish.github.io - ã¯ã˜ã‚ã« - - å…¨ä½“æ„Ÿ - - æ©Ÿæ¢°å¦ç¿’è¶³å›žã‚Šé–¢é€£ã®ã‚¯ãƒ¬ãƒ¼ãƒˆ - Jupyter Notebook Numpy/Scipy Pandas ç”»åƒå‡¦ç† å½¢æ…‹ç´ è§£æž/tokenize - scikit-learnçš„ãªã‚„ã¤ - å„ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã¨ç‰¹å¾´æ¯”è¼ƒ - Gradient Boosting - XGBoos
n_knuu 2021/01/24
Rust

æ©Ÿæ¢°å¦ç¿’

ãƒ©ã‚¤ãƒ–ãƒ©ãƒª
ãƒªãƒ³ã‚¯
What software engineers can bring to machine learning
There are many open questions in machine learning that are only going to be solved through breakthroughs in research. Those probl ems are down to data scientists and researchers. At the same time, there are many challenges within production machine learning that closely parallel challenges in software engineering â€” probl ems weâ€™ve spent decades solving. What we need, and what weâ€™re seeing happen, is
n_knuu 2020/09/22
ã‚½ãƒ•ãƒˆã‚¦ã‚§ã‚¢

æ©Ÿæ¢°å¦ç¿’
ãƒªãƒ³ã‚¯
å…¨ã¦ã®æ©Ÿæ¢°å¦ç¿’ã®è«–æ–‡ã¯æ–°ã—ã„ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã‚’ææ¡ˆã—ã¦ã„ã‚‹ã®ã§ã™ã‹ï¼Ÿ
å›žç” (2ä»¶ä¸ã®1ä»¶ç›®) æ‚²ã—ã„ã“ã¨ã«ãã®é€šã‚Šã§ã™ã€‚ãã—ã¦ã“ã‚Œã“ããŒã“ã®åˆ†é‡Žã®æœ€ã‚‚æ·±ã„å•é¡Œã§ã™ã€‚ç§ã®æŽ¨å®šã§ã¯ã€æ©Ÿæ¢°å¦ç¿’ã§ã¯æ¯Žå¹´10,000ä»¥ä¸Šã®è«–æ–‡ãŒç™ºè¡¨ã•ã‚Œã¦ã„ã¾ã™(ä¸€æ—¥ã«30æœ¬ç¨‹åº¦)ã€‚2020å¹´ã¯ç§ãŒæ©Ÿæ¢°å¦ç¿’ã§æ´»ç™ºã«è«–æ–‡ã‚’ç™ºè¡¨ã—ã¦ã‹ã‚‰35å¹´ç›®ã®å¹´ãªã®ã§ã€ç§ã‚‚æ©Ÿæ¢°å¦ç¿’ã®ç ”ç©¶è€…ã¨åŒã˜ã‚ˆã†ã«ã“ã®ç½ªã‚’çŠ¯ã—ã¦ã„ã¾ã™ã€‚ ãªãœãã‚ŒãŒå•é¡Œãªã®ã‹ã‚’ç†è§£ã—ã¦ã¿ã¾ã—ã‚‡ã†ã€‚è¦å‘Šï¼šä»¥ä¸‹ã®è°è«–ã¯ã€MLã®ç ”ç©¶è€…ã‚„å®Ÿè·µè€…ã¨ã—ã¦ã€ã‚ãªãŸã«æ·±ã„ä¸å®‰ã‚’ä¸Žãˆã¦ã—ã¾ã†ã‹ã‚‚ã—ã‚Œã¾ã›ã‚“ã€‚ç§ã®æŽ¨è«–ã«å°‘ã—ã§ã‚‚æˆ‘æ…¢ã—ã¦ã„ãŸã ã‘ã‚Œã°ã€ç§ãŒå¾—ã‚‰ã‚Œãªã‹ã£ãŸå¤§ããªåˆ©ç›Šã‚’å¾—ã‚‹ã“ã¨ãŒã§ãã‚‹ã‹ã‚‚ã—ã‚Œã¾ã›ã‚“ã€‚ç§ã¯40å¹´é–“æ©Ÿæ¢°å¦ç¿’ã«ã¤ã„ã¦è€ƒãˆã¦ãã¾ã—...
n_knuu 2020/09/15
æ©Ÿæ¢°å¦ç¿’

quora
ãƒªãƒ³ã‚¯
å‹¾é…ãƒ–ãƒ¼ã‚¹ãƒ†ã‚£ãƒ³ã‚°ã®åŸºç¤Žã¨æœ€æ–°ã®å‹•å‘ (MIRU2020 Tutorial)
Gradient Boostingã¯è¿‘å¹´Kaggleãªã©ã®ã‚³ãƒ³ãƒšãƒ†ã‚£ã‚·ãƒ§ãƒ³ã§æ³¨ç›®ã‚’é›†ã‚ã¦ã„ã‚‹åˆ†é¡žã‚„å›žå¸°å•é¡Œã«å¯¾ã™ã‚‹ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã®ä¸€ã¤ã§ã‚ã‚‹ã€‚XGBoost, LightGBM, CatBoostãªã©ãŒæœ‰åã§ã¯ã‚ã‚‹ãŒã€ãã‚Œã‚‰ã‚’åœŸå°ã«ã—ã¦è¿‘å¹´ã¯DeepGBMã‚„NGBoostã¨ã„ã£ãŸæ–°è¦ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã®ç™»å ´ã€æ±ŽåŒ–ç†è«–è§£æžã®é€²å±•ã€ãƒ¢ãƒ‡ãƒ«è§£é‡ˆæ€§ã®ä»˜ä¸Žæ–¹æ³•ã®å¤šæ§˜åŒ–ãªã©ã€ç†è«–ã‹ã‚‰å¿œç”¨ã¾ã§å¤šå²ã«ã‚ãŸã‚‹ç ”ç©¶ãŒè¡Œã‚ã‚Œã¦ã„ã‚‹ã€‚æœ¬ãƒãƒ¥ãƒ¼ãƒˆãƒªã‚¢ãƒ«ã§ã¯ã€Gradient Boostingã«é–¢ã™ã‚‹è¿‘å¹´ã®ç ”ç©¶å‹•å‘ã‚„ãƒ†ã‚¯ãƒ‹ãƒƒã‚¯ã‚’ã€ãã‚Œã‚‰ã®ç¤¾ä¼šå®Ÿè£…ã¾ã§ã‚’è¦‹æ®ãˆãªãŒã‚‰ç´¹ä»‹ã—ã¦ã„ãã€‚ Read less
n_knuu 2020/08/03
å‹¾é…ãƒ–ãƒ¼ã‚¹ãƒ†ã‚£ãƒ³ã‚°

æ©Ÿæ¢°å¦ç¿’
ãƒªãƒ³ã‚¯
Machine Learning Summer School 2020
n_knuu 2020/07/07
æ©Ÿæ¢°å¦ç¿’
ãƒªãƒ³ã‚¯
Learning DAGs with Continuous Optimization
n_knuu 2020/04/23
æ©Ÿæ¢°å¦ç¿’

ã‚°ãƒ©ãƒ•

ç ”ç©¶
ãƒªãƒ³ã‚¯
Sign in - Google Accounts
Not your computer? Use a private browsing window to sign in. Learn more about using Guest mode
n_knuu 2019/10/13
æ©Ÿæ¢°å¦ç¿’

è¡Œåˆ—åˆ†è§£
ãƒªãƒ³ã‚¯
Papers with Code - Browse the State-of-the-Art in Machine Learning
Browse State-of-the-Art 11,969 benchmarks Â Â 5,160 tasks 144,007 papers with code
n_knuu 2019/02/23
æ©Ÿæ¢°å¦ç¿’
ãƒªãƒ³ã‚¯
ã€PyCon JP 2018ã€‘ãƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ã‚¹ã§ç«¶é¦¬ã®çµæžœã‚’äºˆæ¸¬ã™ã‚‹ã€æ©Ÿæ¢°å¦ç¿’ãƒ¢ãƒ‡ãƒ«ã®æ§‹ç¯‰ãƒ—ãƒã‚»ã‚¹
2018å¹´9æœˆ17æ—¥ã‹ã‚‰18æ—¥ã«ã‹ã‘ã¦ã€æ—¥æœ¬æœ€å¤§ã®Pythonã®ç¥å…¸ã€PyCon JP 2018ãŒé–‹å‚¬ã•ã‚Œã¾ã—ãŸã€‚ã€Œã²ã‚ãŒã‚‹Pythonã€ã‚’ã‚ãƒ£ãƒƒãƒã‚³ãƒ”ãƒ¼ã«ã€æ—¥æœ¬ã ã‘ã§ãªãä¸–ç•Œå„åœ°ã‹ã‚‰Python ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ãŸã¡ãŒä¸€å ‚ã«ä¼šã—ã€æ§˜ã€…ãªçŸ¥è¦‹ã‚’å…±æœ‰ã—ã¾ã™ã€‚ãƒ—ãƒ¬ã‚¼ãƒ³ãƒ†ãƒ¼ã‚·ãƒ§ãƒ³ã€Œå®Ÿè·µãƒ»ç«¶é¦¬ãƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ã‚¹ã€ã«ç™»å£‡ã—ãŸã®ã¯ã€è²«äº•é§¿æ°ã€‚äººå·¥çŸ¥èƒ½ï¼ˆAIï¼‰ã«ã‚ˆã‚‹ç«¶é¦¬ã®äºˆæ¸¬ãƒ¢ãƒ‡ãƒ«æ§‹ç¯‰ã«ãŠã„ã¦ç™ºè¦‹ã—ãŸã€å®Ÿè·µçš„ãªãƒŽã‚¦ãƒã‚¦ã‚’ç´¹ä»‹ã—ã¾ã™ã€‚è¬›æ¼”è³‡æ–™ã¯ã“ã¡ã‚‰ ç«¶é¦¬Ã—ãƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ã‚¹ è²«äº•é§¿æ°ï¼ˆä»¥ä¸‹ã€è²«äº•ï¼‰ï¼šã€Œå®Ÿè·µãƒ»ç«¶é¦¬ãƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ã‚¹ã€ã¨é¡Œã—ã¾ã—ã¦ã€AlphaImpactã®è²«äº•ãŒç™ºè¡¨ã—ã¾ã™ã€‚ä»Šå›žã®è³‡æ–™ã¯å¾Œæ—¥å…¬é–‹ã™ã‚‹äºˆå®šãªã®ã§ã€å†™çœŸã¯æ’®ã‚‰ãªãã¦ã‚‚å¤§ä¸ˆå¤«ã§ã™ã€‚ ã¾ãšè‡ªå·±ç´¹ä»‹ã§ã™ãŒã€è²«äº•é§¿ã¨ç”³ã—ã¾ã™ã€‚ å°‚é–€ã¯æ©Ÿæ¢°å¦ç¿’ã‚’ã‚„ã£ã¦ã„ã¾ã—ã¦ã€ä»•äº‹ã¯ã€ã¾ãšæœ¬æ¥ã§Fringe81ã¨ã„ã†ä¼šç¤¾ã§ã‚¢ãƒ‰ãƒ†ãƒƒã‚¯ã¨HRãƒ†ãƒƒã‚¯é ˜åŸŸã®ãƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ãƒ†ã‚£ã‚¹ãƒˆ
n_knuu 2018/09/28
æ©Ÿæ¢°å¦ç¿’

ç«¶é¦¬
ãƒªãƒ³ã‚¯
Deep Reinforcement Learning Doesn't Work Yet
In a world where everyone has opinions, one man...also has opinions June 24, 2018 note: If you want to cite an example from the post, please cite the paper which that example came from. If you want to cite the post as a whole, you can use the following BibTeX: @misc{rlbl ogpost, title={Deep Reinforcement Learning Doesn't Work Yet}, author={Irpan, Alex}, howpublished={\url{https://www.alexirpan.com/
n_knuu 2018/02/18
æ©Ÿæ¢°å¦ç¿’

å¼·åŒ–å¦ç¿’

DQN

æ·±å±¤å¦ç¿’
ãƒªãƒ³ã‚¯
æ•µå¯¾çš„æ‘‚å‹•å…¥é–€
æ•°ãƒ¶æœˆå’è«–ã§å–ã‚Šçµ„ã‚“ã ç¨‹åº¦ã®çŸ¥è˜ã§æç¸®ã§ã™ãŒã€æœ¬è¨˜äº‹ã§ã¯adversarial perturbationã®ç ”ç©¶ã®å…¨ä½“åƒãŒæŠŠæ¡ã§ãã‚‹ã“ã¨ã‚’ç›®æŒ‡ã—ã¾ã™ã€‚æ•µå¯¾çš„æ‘‚å‹•ã¯adversarial perturbationã®ç›´è¨³ã§(å¿…è¦ã«ã‹ã‚‰ã‚Œã¦)åƒ•ãŒå‹æ‰‹ã«å‘¼ã‚“ã§ã„ã‚‹ã ã‘ã®ç”¨èªžã§ã™ã€‚ã“ã®è¨˜äº‹ã¯ãªã‚‹ã¹ãçŸãã¾ã¨ã‚ã¾ã™ãŒã€éœ€è¦ãŒã‚ã‚Œã°ã‚‚ã†å°‘ã—é•·ã„è¨˜äº‹ã‚‚æ›¸ãã‹ã‚‚ã—ã‚Œãªã„ã§ã™ã€‚ index æ•µå¯¾çš„æ‘‚å‹•ã¨ã¯ æ•µå¯¾çš„æ‘‚å‹•ã‚’ä½œã‚‹ æ•µå¯¾çš„æ‘‚å‹•ã‹ã‚‰å®ˆã‚‹ æ•µå¯¾çš„æ‘‚å‹•ã¨ã¯ â€“ æœ‰åãªã®ã¯ä¸‹ã®å›³ã§ã™ãã€‚å‡ºå…¸:Explaining and Harnessing Adversarial Examples. æ¦‚è¦ã‚’ã–ã£ãã‚Šèª¬æ˜Žã™ã‚‹ã¨ã€äººã®ç›®ã«ã¯è¦‹ãˆãªã„ãƒŽã‚¤ã‚ºã‚’åŠ ãˆã‚‹ã“ã¨ã§ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã®å‡ºåŠ›ã‚’ã‚³ãƒ³ãƒˆãƒãƒ¼ãƒ«ã§ãã¦ã‚»ã‚ãƒ¥ãƒªãƒ†ã‚£ä¸Šã‚„ã°ã„ã¨ã„ã†è©±ã§ã™ã€‚ ã‚¬ãƒ¼ãƒ‰ãƒ¬ãƒ¼ãƒ«ã«è‡ªå‹•é‹è»¢è€…ã‚’çªã£è¾¼ã¾ã›ã‚‹ã¨ã‹stop signã‚’ç„¡è¦–ã•ã›ã‚‹ã¨ã‹è‡ªå‹•æ¤œçŸ¥ã‚·ã‚¹ãƒ†
n_knuu 2018/02/17
æ©Ÿæ¢°å¦ç¿’

GAN

æ·±å±¤å¦ç¿’
ãƒªãƒ³ã‚¯
ã‚µãƒ–ã‚«ãƒ«ã®ãŸã‚ã®word2vec | BLOG - DeNA Engineering
ã¯ã˜ã‚ã« AIã‚·ã‚¹ãƒ†ãƒ éƒ¨AIç ”ç©¶é–‹ç™ºã‚°ãƒ«ãƒ¼ãƒ— ã‚¢ãƒ«ãƒã‚¤ãƒˆã®äº”ååµã§ã™ã€‚ï¼ˆ @bonprosoft, ãƒãƒ¼ãƒˆãƒ•ã‚©ãƒªã‚ªï¼šhttp://vbcpp.net/about/ ï¼‰ ç¾åœ¨ã€æ±åŒ—å¤§å¦å¤§å¦é™¢ã®ä¿®å£«1å¹´ã§ã€å¤§å¦é™¢ã§ã¯ï¼ˆè‡ªç„¶è¨€èªžã§ã¯ãªãï¼‰é«˜é€Ÿãªæ–‡å—åˆ—å‡¦ç†ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã«é–¢ã™ã‚‹ç ”ç©¶ã‚’è¡Œã£ã¦ã„ã¾ã™ã€‚ ç§ã¯2017å¹´9æœˆä¸Šæ—¬ã‹ã‚‰3é€±é–“ã»ã©ã€ã‚¢ãƒ«ãƒã‚¤ãƒˆå…¼ã‚¤ãƒ³ã‚¿ãƒ¼ãƒ³ã¨ã—ã¦ãƒãƒƒã‚«ãƒ‰ãƒ¼ãƒ«ãƒãƒ¼ãƒ å†…ã®NLPã®ã‚¿ã‚¹ã‚¯ã«å–ã‚Šçµ„ã‚“ã§ã„ã¾ã—ãŸã€‚ ãã®å¾Œã¯ã‚¢ãƒ«ãƒã‚¤ãƒˆã¨ã—ã¦ã€æœŸé–“ä¸ã«ã§ããªã‹ã£ãŸè¿½åŠ å®Ÿé¨“ã¨å®Ÿéš›ã®è£½å“ã¸ã®é©ç”¨ã«å–ã‚Šçµ„ã‚“ã§ã„ã¾ã™ã€‚ å–ã‚Šçµ„ã‚“ã ã‚¿ã‚¹ã‚¯ çªç„¶ã§ã™ãŒã€ã¿ãªã•ã¾ã€ãƒãƒƒã‚«ãƒ‰ãƒ¼ãƒ«ã¯ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã•ã‚Œã¦ã„ã¾ã™ã‹ï¼Ÿ ãƒãƒƒã‚«ãƒ‰ãƒ¼ãƒ«ã¯ã€ä¸»ã«ã‚µãƒ–ã‚«ãƒ«ãƒãƒ£ãƒ¼ã«é–¢ã™ã‚‹è¨˜äº‹ã«ç‰¹åŒ–ã—ãŸã€ãƒ‹ãƒ¥ãƒ¼ã‚¹ã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ã§ã™ã€‚ ã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³å†…ã®ãƒ¦ãƒ¼ã‚¶ãƒ¼ã®ã‚¯ãƒªãƒƒã‚¯ã‚„ã€Œãƒ›ã‚·ã‚¤/ã‚¤ãƒ©ãƒŠã‚¤ã€ãªã©ã®ã‚¢ã‚¯ã‚·ãƒ§ãƒ³ã‚’é€šã—ã¦ã€ãƒãƒƒã‚«ãƒ‰ãƒ¼ãƒ«ãŒãƒ¦ãƒ¼ã‚¶ãƒ¼ã®å¥½ã¿
n_knuu 2018/01/13
DeNA

word2vec

æ©Ÿæ¢°å¦ç¿’
ãƒªãƒ³ã‚¯
Vol.31.No.5(2016/9)å¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œ | äººå·¥çŸ¥èƒ½å¦ä¼š (The Japanese Society for Artificial Intelligence)
å°å®®å±±ç´”å¹³ï¼ˆæ±äº¬å¤§å¦ ç”Ÿç”£æŠ€è¡“ç ”ç©¶æ‰€ï¼‰ ã¯ã˜ã‚ã«å¤šè…•ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œï¼ˆãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œ, multi-armed bandit probl emï¼‰ã¯ã€è¤‡æ•°ã®ã‚¢ãƒ¼ãƒ ã¨å‘¼ã°ã‚Œã‚‹å€™è£œã‹ã‚‰æœ€ã‚‚è‰¯ã„ã‚‚ã®ã‚’é€æ¬¡çš„ã«æŽ¢ã™å•é¡Œã§ã‚ã‚‹ã€‚ ã‚¢ãƒ¼ãƒ ã¨ã„ã†å¥‡å¦™ãªå˜èªžã¯ã“ã®å•é¡Œã®ã‚‚ã¨ã«ãªã£ãŸã‚¹ãƒãƒƒãƒˆãƒžã‚·ãƒ³ï¼ˆãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆãƒžã‚·ãƒ³ï¼‰ã®æ¯”å–©ã‹ã‚‰æ¥ã¦ã„ã‚‹ã€‚ äºˆæ¸¬è€…ã¯ã„ãã¤ã‹ã®ã‚¹ãƒãƒƒãƒˆãƒžã‚·ãƒ³ã‚’ä¸Žãˆã‚‰ã‚Œã€ãã‚Œãžã‚Œã®ã‚¹ãƒãƒƒãƒˆãƒžã‚·ãƒ³ã‚’å¼•ãã¨å¯¾å¿œã—ãŸå ±é…¬ãŒå¾—ã‚‰ã‚Œã‚‹ã€‚ç¹°ã‚Šè¿”ã™è©¦è¡Œï¼ˆã‚¢ãƒ¼ãƒ ã®é¸æŠžï¼‰ã‚’é€šã˜ã¦å¾—ã‚‰ã‚Œã‚‹å ±é…¬ã‚’æœ€å¤§åŒ–ã™ã‚‹ã®ãŒã€äºˆæ¸¬è€…ã®ç›®æ¨™ã§ã‚ã‚‹ã€‚ å ±é…¬ã‚’æœ€å¤§åŒ–ã™ã‚‹ã¨ã„ã†ç‚¹ã§ã€ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œã¯å¼·åŒ–å¦ç¿’ã®ã‚«ãƒ†ã‚´ãƒªã«å±žã™ã‚‹ã€‚ å®Ÿéš›ã€Suttonã‚‰ã«ã‚ˆã‚‹å¼·åŒ–å¦ç¿’ã®ã‚¯ãƒ©ã‚·ãƒƒã‚¯ãªæ•™ç§‘æ›¸[2]ã§ã‚‚ã€ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œã¯å°ç¯€ã‚’å‰²ãèª¬æ˜Žã•ã‚Œã¦ã„ã‚‹ã€‚ ã‚¢ãƒ¼ãƒ ã¯ã€å¼·åŒ–å¦ç¿’ã®åˆ†é‡Žã§ã¯ã‚¢ã‚¯ã‚·ãƒ§ãƒ³ã‚‚ã—ãã¯ã‚³ãƒ³ãƒˆãƒãƒ¼ãƒ«ã¨å‘¼ã°ã‚Œã‚‹ã“ã¨ãŒã‚ã‚‹ã€‚ ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œã®äºˆ
n_knuu 2017/11/22
å¤šæ¤€ãƒãƒ³ãƒ‡ã‚£ãƒƒãƒˆå•é¡Œ

æ©Ÿæ¢°å¦ç¿’
ãƒªãƒ³ã‚¯
ã‚ã‹ã‚‹LSTM ï½ž æœ€è¿‘ã®å‹•å‘ã¨å…±ã« - Qiita
Machine Learning Advent Calendar 2015 ç¬¬14æ—¥ã§ã™ã€‚åŽ»å¹´ã®Advent Calendarã§å‘³ã‚’ã—ã‚ãŸã®ã§ä»Šå¹´ã‚‚æ›¸ãã¾ã—ãŸã€‚è³ªå•ã€æŒ‡æ‘˜ç‰æ“è¿Žã§ã™ã€‚ ã“ã®è¨˜äº‹ã®ç›®çš„ ã“ã“2ï½ž3å¹´ã®Deep Learningãƒ–ãƒ¼ãƒ ã«åˆã‚ã›ã¦ã€ãƒªã‚«ãƒ¬ãƒ³ãƒˆãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã®ä¸€ç¨®ã§ã‚ã‚‹Long short-term memory(LSTM)ã®å˜åœ¨æ„ŸãŒå¢—ã—ã¦ãã¦ã„ã¾ã™ã€‚LSTMã¯ç¾åœ¨Google Voiceã®åŸºç›¤æŠ€è¡“ã‚’ã¯ã˜ã‚ã¨ã—ãŸæœ€å…ˆç«¯ã®åˆ†é‡Žã§ã‚‚åˆ©ç”¨ã•ã‚Œã¦ã„ã¾ã™ãŒã€ãã®ç™»å ´ã¯1995å¹´ã¨ãã®ã‚¤ãƒ¡ãƒ¼ã‚¸ã¨ã¯è£è…¹ã«æ´å²ã®ã‚ã‚‹ãƒ¢ãƒ‡ãƒ«ã§ã‚‚ã‚ã‚Šã¾ã™ã€‚ã¨ã“ã‚ãŒLSTMã«ã¤ã„ã¦ä½¿ã£ã¦ã¿ãŸè¨˜äº‹ã¯ã‚ã‚Œã©ã€è©³ã—ãè§£èª¬ã•ã‚ŒãŸæ—¥æœ¬èªžæ–‡çŒ®ã¯ã‚ã¾ã‚Šè¦‹å½“ãŸã‚‰ãªã„ã€‚ã¯ã¦ã€ã©ã†ã„ã†ã“ã¨ã§ã—ã‚‡ã†ã‹ã€‚ æœ¬è¨˜äº‹ã§ã¯LSTMã®åŸºç¤Žã‚’ã•ã‚‰ã„ã¤ã¤ã€ä¸€ä½“å…¨ä½“LSTMã¨ã¯ä½•è€…ãªã®ã‹ã€LSTMã¯ã©ã“ã«å‘ã‹ã†ã®ã‹ã€ãã®ä¸èº«ã‚’ã¾ã¨ã‚
n_knuu 2017/11/08
æ©Ÿæ¢°å¦ç¿’

æ·±å±¤å¦ç¿’

RNN

LSTM
ãƒªãƒ³ã‚¯
AlphaGo ã¨ AlphaGo Zero ã®è‡ªå·±å¯¾æˆ¦ã«ã‚ˆã‚‹å¦ç¿’éƒ¨åˆ†ã®é•ã„ - ã†ã©ã‚“è¨˜
æµã—èªã¿ã ã¨ã¡ã‚ƒã‚“ã¨åˆ†ã‹ã‚‰ãªã‹ã£ãŸã®ã§ãƒ¡ãƒ¢ã€‚ æº–å‚™(Alpha Go) policy network : ç›¤é¢ã¨ãã®ç‰¹å¾´é‡ã‚’å…¥åŠ›ã¨ã—ã¦å—ã‘å–ã‚Šã€å„ãƒžã‚¹ã«æ‰“ã¤ç¢ºçŽ‡ã‚’è¿”ã™ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã€‚ value network: ç›¤é¢ã¨ãã®ç‰¹å¾´é‡ã‚’å…¥åŠ›ã¨ã—ã¦å—ã‘å–ã‚Šã€ãã®ç›¤é¢ã§ã®å‹çŽ‡ã‚’è¿”ã™ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã€‚ Alpha Go ã§ã¯ã¾ãš policy network ã‚’ãƒ—ãƒã®æ£‹èœãƒ‡ãƒ¼ã‚¿ã‹ã‚‰æ•™å¸«ã‚ã‚Šå¦ç¿’ã§äº‹å‰å¦ç¿’ã•ã›ã€ãã®å¾Œè‡ªå·±å¯¾æˆ¦ã«ã‚ˆã‚‹å¼·åŒ–å¦ç¿’ã«ã‚ˆã£ã¦ã•ã‚‰ã«æ”¹å–„ã•ã›ã¦ã„ãã€‚ Alpha Go ã®å¼·åŒ–å¦ç¿’ãƒ‘ãƒ¼ãƒˆ æ•™å¸«ã‚ã‚Šå¦ç¿’å¾Œã® policy network ã®ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ $\rho_0$ ã‹ã‚‰å¦ç¿’ã‚’ã‚¹ã‚¿ãƒ¼ãƒˆã™ã‚‹ã€‚è‡ªå·±å¯¾æˆ¦ã®çµæžœã‹ã‚‰ policy network ã®ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã¯éšæ™‚æ›´æ–°ã•ã‚Œã¦ã„ãã€‚ãã‚Œã‚‰ã‚’ $\rho_1, \rho_2, \cdots$ ã¨ã™ã‚‹ã€‚$t$ å›žç›®ã®è‡ªå·±å¯¾æˆ¦ã§ã¯ã€ç¾åœ¨ã®ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ $\
n_knuu 2017/10/30
ç ”ç©¶

è«–æ–‡

å¼·åŒ–å¦ç¿’

AlphaGo

æ©Ÿæ¢°å¦ç¿’
ãƒªãƒ³ã‚¯
LIMEã§æ©Ÿæ¢°å¦ç¿’ã®äºˆæ¸¬çµæžœã‚’è§£é‡ˆã—ã¦ã¿ã‚‹ - Qiita
Deleted articles cannot be recovered. Draft of this article would be also deleted. Are you sure you want to delete this article? åˆæŠ•ç¨¿ã§ã™ è¿½è¨˜(20171031) HTMLã‚’å‡ºåŠ›ã™ã‚‹éš›ã®ã‚¨ãƒ³ã‚³ãƒ¼ãƒ‰ãŒasciiã ã£ãŸãŸã‚æ—¥æœ¬èªžãŒæ–‡å—åŒ–ã‘ã—ã¦ã„ã¾ã—ãŸãŒã€UTF-8ã‚’ã‚µãƒãƒ¼ãƒˆã™ã‚‹ã‚ˆã†ã«ãªã£ãŸã‚ˆã†ã§ã™ï¼ ä¸‹è¨˜ã«æ–‡å—åŒ–ã‘å¯¾ç–ã®æ–‡ç« ãŒã‚ã‚Šã¾ã™ãŒã€èªã¿é£›ã°ã—ã¦ã„ãŸã ã‘ã‚Œã°å¹¸ã„ã§ã™ã€‚ å°Žå…¥ ã‚»ã‚¯ã‚·ãƒ¼ãªãƒ‡ãƒ¼ã‚¿ã‚µã‚¤ã‚¨ãƒ³ãƒ†ã‚£ã‚¹ãƒˆã®çš†æ§˜ã«ãŠã‹ã‚Œã¾ã—ã¦ã¯ã€æ—¥ã€…ã®æ¥å‹™ã§æ§˜ã€…ãªæ©Ÿæ¢°å¦ç¿’ã®ãƒ¢ãƒ‡ãƒ«ã‚’æ§‹ç¯‰ã—ã¦ãŠã‚‰ã‚Œã‚‹ã‹ã¨æ€ã„ã¾ã™ã€‚ãã‚Œã¯ä¾‹ãˆã°ã€ã‚ã‚‹ã‚µãƒ¼ãƒ“ã‚¹ã®ãƒ¦ãƒ¼ã‚¶æƒ…å ±ã‚’ç”¨ã„ã¦ã‚³ãƒ³ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã™ã‚‹ã‹ã—ãªã„ã‹ã‚’äºˆæ¸¬ã—ã¦ã„ãŸã‚Šã€ã¾ãŸã‚ã‚‹æ™‚ã¯å¹´åŽã‚’å›žå¸°ã§äºˆæ¸¬ã—ã¦ã„ãŸã‚Šã€ã¾ãŸã‚ã‚‹ã¨ãã¯ãƒ¦ãƒ¼ã‚¶ãŒæŠ•ç¨¿ã—ãŸç”»åƒæƒ…
n_knuu 2017/10/16
æ©Ÿæ¢°å¦ç¿’

è«–æ–‡

è§£é‡ˆæ€§
ãƒªãƒ³ã‚¯
http://aitc.jp/events/20160916-Seika/20160916_%E7%89%B9%E5%88%A5%E8%AC%9B%E6%BC%94_%E6%B0%97%E8%B1%A1%E5%BA%81%E3%81%AB%E3%81%8A%E3%81%91%E3%82%8B%E6%A9%9F%E6%A2%B0%E5%AD%A6%E7%BF%92%E3%81%AE%E5%88%A9%E7%94%A8.pdf
n_knuu 2017/10/16
æ©Ÿæ¢°å¦ç¿’

æ°—è±¡åº
ãƒªãƒ³ã‚¯
Something went wrong...
n_knuu 2017/10/12
æœ¬

æ©Ÿæ¢°å¦ç¿’
ãƒªãƒ³ã‚¯
1 2 3 4 5 6 æ¬¡ã®ãƒšãƒ¼ã‚¸