[B! deep learning] kojosanã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

kojosan id:kojosan

deep learningã«é–¢ã™ã‚‹kojosanã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (11)

${{author_name}}$

{{{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}}

é¡”èªè¨¼æŠ€è¡“: DeepFace ã¨ Pyramid CNN - Preferred Networks Tech Blog
å¾—å±…ã§ã™ã€‚3æœˆä¸‹æ—¬ã¨ã¯æ€ãˆãªã„å¯’ã•ã«å‡ãˆã¦ã„ã¾ã™ã€‚ Facebook ãŒ CVPR2014 ã«æŠ•ç¨¿ã—ã‚¢ã‚¯ã‚»ãƒ—ãƒˆã•ã‚Œã¦ã„ãŸé¡”èªè¨¼ã«é–¢ã™ã‚‹è«–æ–‡ [1] ãŒ MIT Techno logy Review ã«ã¦ç´¹ä»‹ã•ã‚ŒãŸã“ã¨ã§æ³¨ç›®ã‚’é›†ã‚ã¦ã„ã¾ã™ã€‚DeepFace ã¨åä»˜ã‘ã‚‰ã‚ŒãŸæ‰‹æ³•ã§ã€åŒç¤¾ãŒé›†ã‚ãŸ4030äººã®é¡”å†™çœŸ440ä¸‡æžšã‚’ç”¨ã„ãŸå¤§è¦æ¨¡å¦ç¿’ã«ã‚ˆã£ã¦ã»ã¼äººé–“ä¸¦ã®äººç‰©è˜åˆ¥æ€§èƒ½ã‚’é”æˆã—ã¦ã„ã‚‹ã¨ã„ã†ã“ã¨ã§ã€ãªã‹ãªã‹ã‚ãƒ£ãƒƒãƒãƒ¼ãªè©±é¡Œã§ã™ã€‚ä¸€æ–¹ã€Face++ ã¨ã„ã†é¡”èªè¨¼ãƒ»åˆ†é¡žã®ãƒ—ãƒ©ãƒƒãƒˆãƒ•ã‚©ãƒ¼ãƒ ã‚’å±•é–‹ã™ã‚‹ Megviiç¤¾ ãŒã¤ã„å…ˆæ—¥å…¬é–‹ã—ãŸãƒ—ãƒ¬ãƒ—ãƒªãƒ³ãƒˆ [2] ã§ã‚‚ DeepFace ã¨åŒç¨‹åº¦ã®æ€§èƒ½ã‚’é”æˆã—ã¦ã„ã¾ã™ã€‚ä»Šæ—¥ã¯ã“ã®2ã¤ã®è«–æ–‡ã‚’è§£èª¬ã—ã¾ã™ã€‚ DeepFace ã®è«–æ–‡ã§ã¯ã€æ¤œå‡ºã•ã‚ŒãŸé¡”çŸ©å½¢ã«å¯¾ã—ã¦ä»¥ä¸‹ã®3ã¤ã®å‡¦ç†ã‚’æ–½ã—ã¦ã„ã¾ã™ã€‚ çŸ©å½¢ã®2æ¬¡å…ƒã‚¢ãƒ©ã‚¤ãƒ³ãƒ¡ãƒ³ãƒˆ 3æ¬¡å…ƒãƒ¢ãƒ‡ãƒ«ã‚’ç”¨ã„ãŸ out-of-plan
kojosan 2014/03/21
machine learning

deep learning

é¡”èªè˜
ãƒªãƒ³ã‚¯
è‡ªç„¶è¨€èªžå‡¦ç†åˆ†é‡Žã«ãŠã‘ã‚‹ ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°ã®ç¾çŠ¶
â¾ƒà žç„¶â¾”à¢ èªžå‡¦ç†à¾žåˆ†é‡Žã«ãŠã‘ã‚‹ ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°ã®ç¾çŠ¶ æ¸¡é‚‰â€ƒÌ€é™½å¤ªéƒŽà¾’ æ±åŒ—à¾–â¼¤Ö’å¦â¼¤Ö’å¦é™¢æƒ…å ±ç§‘å¦ç ”ç©¶ç§‘ IBIS2013 ä¼ç”»ã‚»ãƒƒã‚·ãƒ§ãƒ³2ï¼šãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚° 2013/11/12 NLPã«ãŠã‘ã‚‹ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚° 2 â¾”à¢ èªžãƒ¢ãƒ‡ãƒ« ã®æ§‹ç¯‰ â¾”à¢ èªžã®æ§‹æˆæ€§ ã®ãƒ¢ãƒ‡ãƒ«åŒ– æ§‹æˆçš„æ„å‘³è«–à¾” â¾”à¢ èªžè§£æž ï¼ˆæ§‹é€ äºˆæ¸¬ï¼‰ Recursive Neural Networks Autoencoders (Socher et al., 2011, 2012, 2013) RBM (Minh and Hinton 2007) Feed-forward Deep NN (Bengio et al., 2003, Arisoy et al., 2012) Recurrent NN (Mikolov et al., 2010) (Wang and Manning 2013) (Mansur et al.,
kojosan 2013/11/13
NLP

deep learning

machine learning
ãƒªãƒ³ã‚¯
Google TechTalks
Google Tech Talks is a grass-roots program at Google for sharing information of interest to the technical community. At its best, it's part of an ongoing di...
kojosan 2013/07/24
deep learning
ãƒªãƒ³ã‚¯
Deep Learning Tutorials â€” DeepLearning 0.1 documentation
Deep Learning TutorialsÂ¶ Deep Learning is a new area of Machine Learning research, which has been introduced with the objective of moving Machine Learning closer to one of its original goals: Artificial Intelligence. See these course notes for a brief introduction to Machine Learning for AI and an introduction to Deep Learning algorithms. Deep Learning is about learning multiple levels of represen
kojosan 2012/12/09
deep learning
ãƒªãƒ³ã‚¯
å¤§è„³çš®è³ªã¨ deep learning ã®é¡žä¼¼ç‚¹ã¨ç›¸é•ç‚¹
è„³ã¨deep learning ã®ã‚¢ãƒ¼ã‚ãƒ†ã‚¯ãƒãƒ£ã«ã¯å…±é€šã®ç‰¹å¾´ãŒå¤šãã‚ã‚Šã¾ã™ãŒã€ è„³ã«ã¯ã‚ã‚‹ã®ã«ç¾åœ¨(2012å¹´æ™‚ç‚¹)ã® deep learning ã«ã¯ãªã„é‡è¦ãªç‰¹å¾´ã‚‚ã‚ã‚Šã¾ã™ã€‚ ãã®ä¸ã«ã¯ deep learning ã®æ€§èƒ½ã‚’ã•ã‚‰ã«å‘ä¸Šã•ã›ã‚‹ æœ‰æœ›ãªãƒ’ãƒ³ãƒˆãŒå«ã¾ã‚Œã¦ã„ã‚‹ã®ã§ã¯ãªã„ã‹ã¨æ€ã„ã¾ã™ã€‚ ãã“ã§ã€å¤§è„³çš®è³ªã¨ deep learning ã®é¡žä¼¼ç‚¹ã¨ç›¸é•ç‚¹ã‚’ç°¡å˜ã«ã¾ã¨ã‚ã¦ã¿ã¾ã—ãŸã€‚ ç‰¹ã«ã€Œè„³ã¯ä¸Šã®å±¤ã»ã©ç™ºç«ãŒã‚¹ãƒ‘ãƒ¼ã‚¹ã€ã€Œè„³ã¯ã‚ã¾ã‚Šæ·±ããªãã‚€ã—ã‚æ¨ªã«åºƒã„ã€ ã€Œè„³ã§ã¯é ˜é‡Žã”ã¨ã«å¼·ã„å€‹æ€§ãŒã‚ã‚‹ã€ã¨ã„ã£ãŸç‰¹å¾´ã¯ã€ é‡è¦ãªã®ã§ã¯ãªã„ã‹ã¨æ€ã„ã¾ã™ã€‚ â—† å¤§è„³çš®è³ªã«è¦‹ã‚‰ã‚Œã‚‹ã€Œæ·±ã„ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã€ å¤§è„³çš®è³ªã®è¦–è¦šé‡Žï¼ˆè…¹å´çµŒè·¯ã¨èƒŒå´çµŒè·¯ï¼‰ã€è´è¦šé‡Žã€ä½“æ€§æ„Ÿè¦šé‡Žã€é‹å‹•é‡Žã«ã¯ ã€Œæ·±ã„ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã€ã®æ§‹é€ ãŒè¦‹ã‚‰ã‚Œã‚‹ã€‚ ã“ã‚Œã‚‰ã®é ˜åŸŸã«ãŠã‘ã‚‹ä¸»ãªé ˜é‡Žã®éšŽå±¤æ§‹é€ ã‚’ [Felleman and Essen 1991]
kojosan 2012/12/07
deep learning
ãƒªãƒ³ã‚¯
Deep Learning
Deep Learning is a new area of Machine Learning research, which has been introduced with the objective of moving Machine Learning closer to one of its original goals: Artificial Intelligence. This website is intended to host a variety of resources and pointers to information about Deep Learning. In these pages you will find a reading list, links to software, datasets, a list of deep learning resea
kojosan 2012/11/26
deep learning

machine learning
ãƒªãƒ³ã‚¯
Deep Learning
Deep Learning Methods for Vision CVPR 2012 Tutorial 9:00am-5:30pm, Sunday June 17th, Ballroom D (Full day) Rob Fergus (NYU), Honglak Lee (Michigan), Marc'Aurelio Ranzato (Google) Ruslan Salakhutdinov (Toronto), Graham Taylor (Guelph), Kai Yu (Baidu) Overview Hand-designed features such as SIFT and HOG underpin many successful object recognition approaches. However, these only capture low-level edg
kojosan 2012/11/26
deep learning

machine learning
ãƒªãƒ³ã‚¯
Restricted Boltzmann Machineã®å¦ç¿’æ‰‹æ³•ã«ã¤ã„ã¦ã®ç°¡å˜ãªã¾ã¨ã‚
è¿‘å¹´ã®æ©Ÿæ¢°å¦ç¿’ã§ã¯Deep Learningã¨å‘¼ã°ã‚Œã‚‹åˆ†é‡ŽãŒä¸€ä¸–ã‚’é¢¨é¡ã—ã¦ã„ã¾ã™ï¼Žã‚³ãƒ³ãƒ”ãƒ¥ãƒ¼ã‚¿ãƒ“ã‚¸ãƒ§ãƒ³ã‚„è‡ªç„¶è¨€èªžå‡¦ç†ï¼ŒéŸ³å£°èªè˜ãªã©ã®åˆ†é‡Žã§ã¯ä½•ã‚‰ã‹ã®å•é¡Œã‚’è§£ã“ã†ã¨ã—ãŸéš›ã«ï¼Œã¾ãšå¯¾è±¡ã®å…¥åŠ›ãƒ‡ãƒ¼ã‚¿ã‹ã‚‰SIFTã‚„ã‚±ãƒ—ã‚¹ãƒˆãƒ©ãƒ ã¨ã„ã£ãŸä½•ã‚‰ã‹ã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã‚’ç”¨ã„ã¦ç‰¹å¾´ãƒ™ã‚¯ãƒˆãƒ«ã‚’æŠ½å‡ºã—ï¼Œã”ã‚Šã”ã‚Šã¨åˆ¤åˆ¥ã—ã¦ã„ãã¨ã„ã£ãŸæµã‚ŒãŒä¸€èˆ¬çš„ã§ã™ï¼Žã—ã‹ã—ï¼Œãã®ç‰¹å¾´ãƒ™ã‚¯ãƒˆãƒ«ã‚’ç”Ÿæˆã™ã‚‹ã¨ã„ã†ç”Ÿã®ãƒ‡ãƒ¼ã‚¿ã‹ã‚‰æœ¬è³ªã¨ãªã‚‹éƒ¨åˆ†ã‚’æŠ½å‡ºã™ã‚‹ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ è‡ªä½“ã¯ç ”ç©¶è€…ãŒä¸€ç”Ÿæ‡¸å‘½è€ƒãˆãªãŒã‚‰ä½œã‚‹ã®ãŒæ™®é€šã§ã—ãŸï¼Ž Deep Learningã®åˆ†é‡Žã§æœ€ã‚‚æœ‰åãªæ‰‹æ³•ã®ä¸€ã¤ã§ã‚ã‚‹Deep Belief Nets(DBN) [Hinton06]ã¯ï¼Œç ”ç©¶è€…ãŒã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã‚’ä½œã‚‹ã®ã§ã¯ãªãï¼Œãã‚Œè‡ªä½“ã‚‚æ©Ÿæ¢°å¦ç¿’ã«ã‚„ã‚‰ã›ã¾ã—ã‚‡ã†ã¨ã„ã†å‹•æ©Ÿã§ç”Ÿã¾ã‚ŒãŸã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã§ã™ï¼ŽDBNã§ã¯ã¾ã‚‹ã§ä¸€æ˜”å‰ã«ã‚„ãŸã‚‰æµè¡Œã£ãŸãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã®ã‚ˆã†ã«å„ãƒŽãƒ¼ãƒ‰ã‚’å±¤çŠ¶ã«é…ç½®ã—ï¼Œã
kojosan 2012/11/24
RBM

Restricted Boltzmann Machine

deep learning
ãƒªãƒ³ã‚¯
ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã®é€†è¥² - Preferred Networks Research & Development
å²¡é‡ŽåŽŸã§ã™ã€‚Deep LearningãŒå„åˆ†é‡Žã®ã‚³ãƒ³ãƒšãƒ†ã‚£ã‚·ãƒ§ãƒ³ã§å„ªå‹ã—è©±é¡Œã«ãªã£ã¦ã„ã¾ã™ã€‚Deep Learningã¯7ã€8æ®µã¨æ·±ã„ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã‚’ä½¿ã†å¦ç¿’æ‰‹æ³•ã§ã™ã€‚ã™ã§ã«ã€ç”»åƒèªè˜ã€éŸ³å£°èªè˜ã€æœ€ã‚‚æœ€è¿‘ã§ã¯åŒ–åˆç‰©ã®æ´»æ€§äºˆæ¸¬ã§å„ªå‹ã—ãŸã‚Šã€æ—¢å˜ãƒ‡ãƒ¼ã‚¿ãƒ»ã‚»ãƒƒãƒˆã§ã®æœ€é«˜ç²¾åº¦ã‚’é”æˆã—ã¦ã„ã¾ã™ã€‚ä»¥ä¸‹ã«å¹¾ã¤ã‹ä¾‹ã‚’ã‚ã’ã¾ã™ã€‚ ç”»åƒèªè˜ LSVRC 2012 [html] Â å„ªå‹ãƒãƒ¼ãƒ ã‚¹ãƒ©ã‚¤ãƒ‰ [pdf], ã¾ã¨ã‚ã‚¹ãƒ©ã‚¤ãƒ‰[pdf] Googleã«ã‚ˆã‚‹å·¨å¤§ãªNeuralNetã‚’åˆ©ç”¨ã—ãŸç”»åƒèªè˜ï¼ˆçŒ«èªè˜ã¨ã—ã¦æœ‰åï¼‰[paper][slide][æ—¥æœ¬èªžè§£èª¬] ã¾ãŸã€å„åˆ†é‡Žã®ãƒˆãƒƒãƒ—ã‚«ãƒ³ãƒ•ã‚¡ãƒ¬ãƒ³ã‚¹ã§Deep Learningã®ãƒãƒ¥ãƒ¼ãƒˆãƒªã‚¢ãƒ«ãŒè¡Œã‚ã‚Œã€ã‚µãƒ¼ãƒ™ã‚¤è«–æ–‡ã‚‚ã„ãã¤ã‹å‡ºã¾ã—ãŸã€‚ãŠãã‚‰ãæ¥å¹´ä»¥é™ã“ã†ã—ãŸè©±ãŒå¢—ãˆã¦ãã‚‹ã“ã¨ãŒè€ƒãˆã‚‰ã‚Œã¾ã™ã€‚ ICML 2012 [pdf] ACL 2012 [pdf] CVPR
kojosan 2012/11/02
deep learning

machine-learning
ãƒªãƒ³ã‚¯
Deep Learning for NLP (without Magic)
kojosan 2012/11/02
deep learning

machine learning

ã‚ã¨ã§èªã‚€
ãƒªãƒ³ã‚¯
representation-learning-tutorial-2012
ICML 2012 Representation Learning Tutorial Yoshua Bengio U. Montreal June 26th, 2012, Edinburgh, Scotland The success of machine learning algorithms generally depends on data representation, and we hypothesize that this is because different representations can entangle and hide more or less the different explanatory factors of variation behind the data.Â Although domain knowledge can be used to he
kojosan 2012/10/26
deep learning

machine learning

ã‚ã¨ã§èªã‚€
ãƒªãƒ³ã‚¯
1