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

hikoma id:hikoma

deep learningã«é–¢ã™ã‚‹hikomaã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (6)

${{author_name}}$

{{{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}}

Deep Learning ã‚’å®Ÿè£…ã™ã‚‹
2. ç›®æ¬¡ â€¢â€¯ Deep Learning ã¨ã¯" â€“â€¯ æ©Ÿæ¢°å¦ç¿’ã«ã¤ã„ã¦" â€“â€¯ å¾“æ¥ã® NN ã¨ã®ã¡ãŒã„" â€“â€¯ Deep Learning ã®ãƒ–ãƒ¬ã‚¤ã‚¯ã‚¹ãƒ«ãƒ¼" â€¢â€¯ dA (Denoising Autoencoders) ã‚’ã†ã”ã‹ã™" â€“â€¯ æ•°ç†ãƒ¢ãƒ‡ãƒ«ã®è§£èª¬" â€“â€¯ Python ã§å®Ÿè£…ã™ã‚‹å‰æº–å‚™" â€“â€¯ ã‚³ãƒ¼ãƒ‰ãƒ¬ãƒ“ãƒ¥ãƒ¼" â€“â€¯ å®Ÿè¡Œçµæžœ" â€¢â€¯ RBM (Restricted Boltzmann Machines) ã‚’ã†ã”ã‹ã™" â€“â€¯ æ•°ç†ãƒ¢ãƒ‡ãƒ«ã®è§£èª¬" â€“â€¯ å®Ÿè¡Œçµæžœ" â€¢â€¯ ã¾ã¨ã‚ 4. Deep Learning ã¨ã¯ â€¢â€¯ å…¥åŠ›ä¿¡å·ã‹ã‚‰ã‚ˆã‚ŠæŠ½è±¡çš„ãªæ¦‚å¿µã‚’å¦ã¶ãƒ»ç‰¹å¾´ã‚’æŠ½å‡ºã™ã‚‹ æ©Ÿæ¢°å¦ç¿’ã®æ‰‹æ³•ã®é›†åˆã§ã™ " â€œãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã¨ã©ã†é•ã†ã®ï¼Ÿâ€! â€¢â€¯ ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã‚’å¤šå±¤ã«ã—ãŸã‚“ã§ã™ " â€œå¾“æ¥ã®ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã¨ä½•ãŒé•ã†ã®ï¼Ÿâ€! â€¢â€¯ ã²ã¨ã¤ã²ã¨ã¤ã®ãƒ¬ã‚¤ãƒ¤ãƒ¼é–“ã§ãƒ‘ãƒ©
hikoma 2015/02/21
æ©Ÿæ¢°å¦ç¿’

deep learning
ãƒªãƒ³ã‚¯
SSSSLIDE
hikoma 2014/10/23
deep learning
ãƒªãƒ³ã‚¯
äººå·¥çŸ¥èƒ½å¦ä¼šèªŒ é€£è¼‰è§£èª¬ã€ŒDeep Learningï¼ˆæ·±å±¤å¦ç¿’ï¼‰ã€
äººå·¥çŸ¥èƒ½å¦ä¼šç›£ä¿®ã€Žæ·±å±¤å¦ç¿’ â€” Deep Learningã€å‡ºç‰ˆã®æ¡ˆå†… æœ¬é€£è¼‰è§£èª¬ã®è¨˜äº‹ã‚’åŠ ç†ãƒ»å†ç·¨é›†ã—ãŸæ›¸ç±ã€Žæ·±å±¤å¦ç¿’ â€” Deep Learningã€ã‚’è¿‘ä»£ç§‘å¦ç¤¾ã‹ã‚‰2015å¹´10æœˆ31æ—¥ã«å‡ºç‰ˆã—ã¾ã—ãŸï¼Ž ç›£ä¿®ï¼šäººå·¥çŸ¥èƒ½å¦ä¼š è‘—è€…ï¼šéº»ç”Ÿ è‹±æ¨¹ï¼Œå®‰ç”° å®—æ¨¹ï¼Œå‰ç”° æ–°ä¸€ï¼Œå²¡é‡ŽåŽŸ å¤§è¼”ï¼Œå²¡è°· è²´ä¹‹ï¼Œä¹…ä¿ é™½å¤ªéƒŽï¼Œãƒœãƒ¬ã‚¬ãƒ©ãƒ»ãƒ€ãƒŒã‚·ã‚« ç·¨é›†ï¼šç¥žå¶Œ æ•å¼˜ å‡ºç‰ˆï¼šè¿‘ä»£ç§‘å¦ç¤¾ å‡ºç‰ˆç¤¾ã®ãƒšãƒ¼ã‚¸ ã‚µãƒãƒ¼ãƒˆãƒšãƒ¼ã‚¸ äººå·¥çŸ¥èƒ½å¦ä¼šèªŒ é€£è¼‰è§£èª¬ã€ŒDeep Learningï¼ˆæ·±å±¤å¦ç¿’ï¼‰ã€ ç¾åœ¨ã§ã¯è¨˜äº‹ã¯ç„¡æ–™ã§å…¬é–‹ã•ã‚Œã¦ã„ã¾ã™ï¼Ž å®‰ç”° å®—æ¨¹ã€Œãƒ‡ã‚£ãƒ¼ãƒ—ãƒœãƒ«ãƒ„ãƒžãƒ³ãƒžã‚·ãƒ³å…¥é–€ : ãƒœãƒ«ãƒ„ãƒžãƒ³ãƒžã‚·ãƒ³å¦ç¿’ã®åŸºç¤Žã€äººå·¥çŸ¥èƒ½å¦ä¼šèªŒ Vol.28 No.3 (2013å¹´5æœˆ) éº»ç”Ÿ è‹±æ¨¹ã€Œå¤šå±¤ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã«ã‚ˆã‚‹æ·±å±¤è¡¨ç¾ã®å¦ç¿’ã€äººå·¥çŸ¥èƒ½å¦ä¼šèªŒ Vol.28 No.4 (2013å¹´7æœˆ) å²¡é‡ŽåŽŸ å¤§è¼”ã€Œå¤§è¦æ¨¡Deep Lear
hikoma 2014/10/20
deep learning
ãƒªãƒ³ã‚¯
è‡ªç„¶è¨€èªžå‡¦ç†ã¾ã‚ã‚Šã®Deep Learningã‚’è‡ªåˆ†ãªã‚Šã«ã¾ã¨ã‚ã¦ã¿ãŸ â€” KiyuHub
è‡ªç„¶è¨€èªžå‡¦ç†ã¾ã‚ã‚Šã®Deep Learningã‚’è‡ªåˆ†ãªã‚Šã«ã¾ã¨ã‚ã¦ã¿ãŸ â€œè‡ªç„¶è¨€èªžå‡¦ç†ã®ãŸã‚ã®Deep Learningâ€ã¨ã„ã†ã‚¹ãƒ©ã‚¤ãƒ‰ã‚’å…¬é–‹ã—ã¾ã—ãŸï¼Ž è‡ªç„¶è¨€èªžå‡¦ç†ã®ãŸã‚ã®Deep Learning from Yuta Kikuchi ã‚«ã‚¸ãƒ¥ã‚¢ãƒ«ãªæ„Ÿã˜ã§è‡ªç„¶è¨€èªžå‡¦ç†ã¾ã‚ã‚Šã®Deep Learningã®è©±é¡Œã‚’ã¾ã¨ã‚ãŸæ„Ÿã˜ã«ãªã£ã¦ã„ã¾ã™ï¼Ž ãã£ã‹ã‘ã¯ï¼Œå‹‰å¼·ä¼šã‚’ã—ã¦ã„ã‚‹ã“ã¨ã‚’çŸ¥ã£ãŸOBã®beatinaniwaã•ã‚“ã«ãŠé¡˜ã„ã•ã‚ŒãŸã“ã¨ã§ï¼Œ æ ªå¼ä¼šç¤¾Gunosyã®å‹‰å¼·ä¼šã®å ´ã§ï¼Œç™ºè¡¨ã®æ©Ÿä¼šã‚’é ‚ãã¾ã—ãŸï¼Ž ãã‚ŒãŒï¼Œ9/11ã§ï¼Œãã®å¾Œ9/26ã«ç ”ç©¶å®¤å†…ã§åŒã˜å†…å®¹ã§ç™ºè¡¨ã—ã¾ã—ãŸï¼Ž ã©ã¡ã‚‰ã‚‚æ€ã£ãŸä»¥ä¸Šã«å¥½è©•ã‚’é ‚ã‘ï¼Œå…¬é–‹ã—ã¦ã¯ã¨é€²ã‚ã¦é ‚ã„ãŸã®ã§ï¼Œå…¬é–‹ã™ã‚‹ã“ã¨ã«ã—ã¾ã—ãŸï¼Ž ã‚‚ã¡ã‚ã‚“é–“é•ã„ãŒå«ã¾ã‚Œã¦ã„ã‚‹å¯èƒ½æ€§ã‚‚å¤šåˆ†ã«ã‚ã‚‹ã®ã§ï¼Žæ°—ã¥ã„ãŸæ–¹ã¯ã”æŒ‡æ‘˜é ‚ã‘ã‚‹ã¨å¹¸ã„ã§ã™ï¼Ž å†…å®¹ã–ã£ãã‚Š å‰åŠã¯ï¼Œãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ã‚’å›³ã‚’ä½¿
hikoma 2014/10/20
deep learning
ãƒªãƒ³ã‚¯
Deep Learning : Bengioå…ˆç”Ÿã®ãŠã™ã™ã‚ãƒ¬ã‚·ãƒ” - a lonely miner
å…ˆæ—¥ï¼Œèº«å†…ã®å‹‰å¼·ä¼š(&ãƒ©ãƒœã®å‹‰å¼·ä¼š)ã§ï¼ŒDeep Learningã«ã¤ã„ã¦ãŠè©±ã—ã¦ãã¾ã—ãŸï¼Žã“ã‚Œã¾ã§èˆˆå‘³ãŒãªã•ãã†ã ã£ãŸã®ã«ä½•æ•…æ€¥ã«ï¼Ÿã¨ã„ã†ã®ã¯ãŠã„ã¦ãŠã„ã¦ï¼Ž ç´¹ä»‹ã—ãŸã®ã¯ï¼ŒDeep Learningã®ç¬¬ä¸€äººè€…ã®ã²ã¨ã‚Šï¼Œ Yoshua Bengioå…ˆç”Ÿè‡ªèº«ãŒåŸ·ç†ã•ã‚ŒãŸï¼Œä»¥ä¸‹ã®è«–æ–‡ï¼Ž Yoshua Bengio, Practical recommendations for gradient-based training of deep architectures, arXiv:1206.5533v2, 2012 ã©ã†ã‚„ã‚‰æ›¸ç±ã®è‰ç¨¿ã®ã‚ˆã†ã§ï¼ŒBengioå…ˆç”Ÿã®é•·å¹´ã®ç ”ç©¶ã§å¾—ã‚‰ã‚ŒãŸã•ã¾ã–ã¾ãªãƒŽã‚¦ãƒã‚¦(æœ€è¿‘ã®æ‰‹æ³•ã¾ã§)ãŒãŽã£ã—ã‚Šè©°ã‚è¾¼ã¾ã‚Œã¦ã„ã¾ã™ï¼Žã™ã”ã„ï¼Ž ä»¥å‰ã‹ã‚‰æ°—ã«ã¯ãªã‚Šã¤ã¤ã‚‚ï¼Œã¡ã‚‡ã£ã¨åˆ†é‡ãŒå¤šã„(30ãƒšãƒ¼ã‚¸ãã‚‰ã„ã‚ã‚‹)ã®ã§ï¼Œãªã‹ãªã‹æ‰‹ã‚’å‡ºã™ã“ã¨ãŒã§ããªã‹ã£ãŸã®ã§ã™ãŒï¼Œã‚ˆã†ã‚„ããƒ¤ãƒ«æ°—ãŒå‡ºã¦ããŸ
hikoma 2014/10/20
deep learning
ãƒªãƒ³ã‚¯
Hello Autoencoder â€” KiyuHub
Hello Autoencoder æœ€è¿‘ï¼Œèº«å†…ã§Deep Learningã‚’é¡Œæã«å«ã‚“ã å‹‰å¼·ä¼šã‚’è¡Œãªã£ã¦ã„ã‚‹ï¼Ž ãƒ¡ã‚¤ãƒ³ã¯å°‚é–€ã§ã‚ã‚‹è‡ªç„¶è¨€èªžå‡¦ç†ã¾ã‚ã‚Šã ãŒï¼Œ ã¨ã‚Šã‚ãˆãšå®Ÿè£…(ã¨ã„ã†ã‚ˆã‚Šnumpy)ã®å°Žå…¥ã«ãªã‚‹ä¸Šï¼Œçµæžœã‚’è¦–è¦šåŒ–ã§ãã‚‹ã“ã¨ã‹ã‚‰ï¼Œ ç”»åƒãƒ‡ãƒ¼ã‚¿ã‚’åˆ©ç”¨ã—ãŸAutoencoderã®å®Ÿè£…ã«ã¤ã„ã¦å–ã‚Šæ‰±ã£ãŸï¼Ž è»½ã„èª¬æ˜Žã¨å…±ã«ã‚³ãƒ¼ãƒ‰ã¨ï¼Œè‰²ã€…ãªçµæžœã‚’ Autoencoder Autoencoderã¨ã¯,Neural Networkã®ç‰¹æ®Šç³»ã§ï¼ŒåŸºæœ¬çš„ã«ã¯ å…¥åŠ›å±¤ã¨å‡ºåŠ›å±¤ã®ãƒ¦ãƒ‹ãƒƒãƒˆæ•°ãŒåŒã˜ã§ã‚ã‚‹ï¼Ž æ•™å¸«ä¿¡å·ã¨ã—ã¦å…¥åŠ›ãã®ã‚‚ã®ã‚’ä¸Žãˆã‚‹ï¼Ž ã¨ã„ã†ç‰¹å¾´ã‚’æŒã¤ï¼Ž å…¥åŠ›ã¨å‡ºåŠ›ãŒå…±ã«4æ¬¡å…ƒã§ï¼Œéš ã‚Œå±¤ãŒ2æ¬¡å…ƒãªAutoencoderã®å›³ Autoencoderã¯ï¼Œå…¥åŠ›ã®æƒ…å ±ã‚’ã‚’ä¸€åº¦éš ã‚Œå±¤ã®ç©ºé–“ã«å†™åƒ(encode) ã—ãŸã‚ã¨ï¼Œ å…ƒã®ä¿¡å·ã‚’å¾©å…ƒ(decode)ã™ã‚‹ã‚ˆã†ãªãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã‚’å¦ç¿’ã™ã‚‹ï¼Ž å›³ã®ã‚ˆã†ã«ï¼Œã‚‚ã—ã‚‚éš ã‚Œ
hikoma 2014/10/20
deep learning
ãƒªãƒ³ã‚¯
1