æ°åŠ
NHKå°äžæ³¢TVã®ãç¬ããªãæ°åŠãã¯ãç®äžå€§å¥œããªTVã·ãªãŒãºã®äžã€ã§ãããææ°ã®ã1+1=2ããããããžãé¢çœãèŠèŽããã www.nhk.jp 19äžçŽã®éãŠãŒã¯ãªãã幟äœåŠç»å Žããã£ããã«ãèªæãšæãããããšã§ãçããªããã°ãªããªãããšæ°åŠçã§ã¯æ°åŠã®åºç€ä»âŠ
ãã€ãã¿ãŒã«æ°åŠã®åé¡ãæµããŠããããã€äž»ããã®ãåããããã£ã¬ã³ãžããŠããå¿çšåé¡ãšã®ããšã ã£ããå ã ã¯æçã®é«æ ¡å ¥è©Šåé¡ããããæŠç¥ã¯æ¬¡ã®éããè©Šéšåé¡ã¯èäœæš©æ³36æ¡ã«ããèäœæš©ã«ããå¶éãåããªãã¯ãã ããã©ã£ãããæããããæ¶ããŸâŠ
ãAIã®èããããšã¯ããããããšããèšèãããããããã©ããç¥ããªããä»ãæãã€ããã åã åããã®ïŒããšååããã®ïŒãã§ã°ã©ããæç»ããåŸã«ãã³ãããéã¿ W1ãW2 ãš ãã€ã¢ã¹ b1ãb2 ã®å€ããExcel ã·ãŒãã«ãŸãšããŠã¿ããExcel ã®è¡šã¯ããŒã ããŒâŠ
ååãšã®éãã¯ãéã¿W1ãšW2ãžã®ä¹æ° weight_init_std ã 0.1 ãšãããšå°ããããããšãšãç¹°ãè¿ãåŠç¿åæ° s_n ã20ãã100ã«å¢ãããããšã®2ç¹ã§ãããç¹°ãè¿ãåŠç¿åæ°ãå¢ãããã®ã¯ã20åã§ã¯æ£è§£çã100ïŒ ã«ãªããªãã£ãããã ã ãã®äžã§ãnp.round()âŠ
ååããã®ïŒãã®æåŸã«æžãããéã¿W1ãW2ããã€ã¢ã¹b1ãb2ã®ã°ã©ãæç»çšããŒã¿ãäžæ¬ããŠæ¡åããã³ãŒãæ¹é ã¯ããã£ããã§ããã #ã³ãŒã4-0import sys, ossys.path.append(os.pardir)import numpy as npfrom common.functions import *from common.gradâŠ
çžå€ãããæè€åº·æ¯ ããŒãããäœãDeep Learning âPythonã§åŠã¶ãã£ãŒãã©ãŒãã³ã°ã®çè«ãšå®è£ ã(O'REILLY) èªè éå®ã®èšäºã§ããããã«ãã¿ãã©ã€å¹æãã«ãªã¹çŸè±¡ã«ããçšåºŠã®ç¥èãšèå³ããã人ãšããããšã§ãããã«èªè ã¯éå®ããããã§ãããããã°ã§âŠ
å¿çšæ°çç§åŠã®è±åœ¢ã§ããæ©æ¢°åŠç¿ãšãããžã£ã³ã«ã§ããããåºæ¬äžã®åºæ¬ã§ããEORåè·¯ãå®çŸããããšããã®ã ãããå è¡ç 究ããªãããããªããšã¯æãã®ã ããèªåã®è¶£å³ã§ãã£ãŠããã®ã§ããããšã«ããã ãžã§ã€ã ãºã»ã°ãªãã¯ãã«ãªã¹âæ°ããç§åŠãã€ããâŠ
ãã¿ãŸããã察象ã¯æè€åº·æ¯ ããŒãããäœãDeep Learning âPythonã§åŠã¶ãã£ãŒãã©ãŒãã³ã°ã®çè«ãšå®è£ ã(O'REILLY) èªè éå®ã®èšäºã«ã€ããæ°çãç®æ±ããé¿ããããæ¥ä»ãããã®ãŒã£ãŠå ¬éããŸããåŒããã°ã§ã¯ããšãã©ãããããããšããããŸãã 2幎åâŠ
æééå®çã«ããã€ãã¿ãŒã®ãèªå·±çŽ¹ä»ãæ¬ã«æ¬¡ã®ããã«è¡šç€ºããŠã¿ãã ããã©ãŒå é£æè²çžãã«ã³ãã«ãã»ã«ã¹ãããŠãã€äººã®åœ±é¿ããªããªã£ãã²ããã³ã²ã³æ°åŠç 究æã¯ã©ãããïŒããã«ãã«ããé£äžããã¯ãã²ããã³ã²ã³ã«æ°åŠã¯ãªããªã£ãŠããŸããŸããã#âŠ
ãã€ãã¿ãŒã«æµããŠããåé¡ã§ãããFF å€ããåŒçšå€±ç€ŒããŸãã é«æ ¡çAãã³â4+2³â10ããšã6ãã©ã¡ãã倧ãããïŒãé«æ ¡çãããŒããšããããã©ã¡ãã3ã§å²ãïŒf(x)=³âxãšããã°äžæ¹ã¯(4,f(4))ãš(10,f(10)ã®2ïŒ1å åç¹ã®y座æšãããïŒä»æ¹f(8)ã ããã»ã»âŠ
3æ¡ã®æ°å195ã 1Ã95 = 19Ã5ïŒ=95ïŒãšããæ§è³ªãæºããããšããããã°ã1995ã19995ã199995âŠãåãæ§è³ªãæºããããšããèšç®ããªããŠããããããšã瀺ãããã ãã®ãããèšç®ããã£ãŠå€§ããæéãããªããšããçªã£èŸŒã¿ã¯é»æ®ºããã 1995ã®å Žåã ã瀺ãã°ãâŠ
ãŸãããŠãããã®åé¡ã®è©±ã§ãã 3æ¡ä»¥äžã®næ¡ã®æŽæ°ã«ãããŠãæäžäœæ¡ãåãåºãäžæ¡ã®æ°åãšèŠãŠãæ®ãã®n-1æ¡ã«æããçµæãšãæäžäœæ¡ã®äžæ¡ã®æ°åããæ®ããn-1æ¡ã®æ°åãšèŠãŠæããçµæããäžèŽããæ°ã«ã¯ã©ããªãã®ããããïŒ å ·äœçã«ã¯111ã222ã3âŠ
èãããŸãšãŸã£ãŠããŸããããéå ±çã«ã ååã®ãšã³ããªãŒã«ã¯ã1æ¥ã§1äž9åè¶ ã®pvãããã ããæè¬ããŠããŸããããããšãããããŸããã ãŸãäœåãã®æ¹ããããã°ã«ãŠèšåãããã ããŸããããã¡ããæè¬ãã€ã€ã倱瀌ããŠãªã³ã¯ã貌ãããŠããã ããŸãïŒâŠ
ç®æ¬¡ ïŒãã®ïŒïŒ ãŸããã ãã§ã«ããŒç¹ãæççµè·¯ãäžãã蚌æ æ°Žå¹³æ¯æ±ãåæ¬ã§ããå Žå æ°Žå¹³æ¯æ±ãå æ¬ã§ããå Žå ïŒãã®ïŒïŒ æ£äºè§åœ¢ã®ã·ã¥ã¿ã€ããŒæš æ£äºè§åœ¢ãå€å¿ãšé ç¹ã§åå²ããäžè§åœ¢ã®ã·ã¥ã¿ã€ããŒæš ïŒãã®ïŒïŒ æ£äºè§åœ¢ã®ç¬¬äžã®äŸ æ£å è§åœ¢ã®âŠ
ç®æ¬¡ ïŒãã®ïŒïŒ ãŸããã ãã§ã«ããŒç¹ãæççµè·¯ãäžãã蚌æ æ°Žå¹³æ¯æ±ãåæ¬ã§ããå Žå æ°Žå¹³æ¯æ±ãå æ¬ã§ããå Žå ïŒãã®ïŒïŒ æ£äºè§åœ¢ã®ã·ã¥ã¿ã€ããŒæš æ£äºè§åœ¢ãå€å¿ãšé ç¹ã§åå²ããäžè§åœ¢ã®ã·ã¥ã¿ã€ããŒæš ïŒãã®ïŒïŒ æ£äºè§åœ¢ã®ç¬¬äžã®äŸ æ£å è§åœ¢ã®âŠ
ç®æ¬¡ ãŸããã ãã§ã«ããŒç¹ãæççµè·¯ãäžãã蚌æ æ°Žå¹³æ¯æ±ãåæ¬ã§ããå Žå æ°Žå¹³æ¯æ±ãå æ¬ã§ããå Žå ïŒãã®ïŒïŒ æ£äºè§åœ¢ã®ã·ã¥ã¿ã€ããŒæš æ£äºè§åœ¢ãå€å¿ãšé ç¹ã§åå²ããäžè§åœ¢ã®ã·ã¥ã¿ã€ããŒæš ïŒãã®ïŒïŒ æ£äºè§åœ¢ã®ç¬¬äžã®äŸ æ£å è§åœ¢ã®äºçš®é¡ã®ã·âŠ
ã¿ãæ¬ã§ããæè€åº·æ¯ ããŒãããäœãDeep Learning âPythonã§åŠã¶ãã£ãŒãã©ãŒãã³ã°ã®çè«ãšå®è£ ã(ãªã©ã€ãªãŒã»ãžã£ãã³)ã§ã¯ãæãExcelã§æ軜ã«è©Šãæ©æ¢°åŠç¿ãã·ãªãŒãºã®1ç« ã§è¿°ã¹ããããªäžæ¬¡äžçåŒã§å®çŸããè«çåè·¯ããããŒã»ãããã³ããšåŒãã§ããâŠ
ORåè·¯ã«é¢ããŠã¯ãåèšäºã®Excelã·ãŒãã«ãããŠã»ã«B13ïœD16ã«å ¥ã£ãŠããççå€è¡šããŒã¿ã®ãã¡ãD14ãšD15ã 1 ã«å€æŽããã ãã§ãããççå€è¡šããŒã¿ãåæå€ãããã«åŸ®å°å€ãªã©ã®æ°å€ããŒã¿ã¯ããããã§ãäžæžããå¯èœã§ããã ä»åãcsvãã¡ã€ã«ã®ãã³ãâŠ
ã³ã³ãã¥ãŒã¿ã®åºç€ãšãªãè«çåè·¯ãæ§æããã«ã¯ãããŸããŸãªæ¹æ³ãããã ããã§ã¯äžçåŒãçšããŠANDåè·¯ãšORåè·¯ãæ§æããŠã¿ãã å€æ° xãy ãšéã¿ W0ãW1ãW2 ã«ã€ããŠã次ã®äžçåŒãèããã W0x + W1y + W2 > 0 ãã®äžçåŒãæç«ããŠããã° 'ç' ããªâŠ
2幎ã»ã©åã«ãæè€åº·æ¯ ããŒãããäœãDeep Learning âPythonã§åŠã¶ãã£ãŒãã©ãŒãã³ã°ã®çè«ãšå®è£ ã(ãªã©ã€ãªãŒãžã£ãã³) ãšããæ¬ããå幎ã»ã©ãããŠèªãã ã ãããã§ãæ©æ¢°åŠç¿ããšãããã®ãã©ããããã®ããèªåãªãã«ç解ã§ããããã«æã£ãã ãã ãâŠ
é£ãããã©ãèããŠã¿ãããç®æ°ãã¿ã°ã¯äœã£ãŠãªããããæ°åŠãã¿ã°ã䜿ããããªãã£ãã ãããŸ@mtchacïŒid:go_kumaïŒããã®ãã¡ãã®èšäºã«ãèšå倱瀌ããŸãã gokumatrix.hateblo.jp äŸãã° 52 ÷ 4 ãçç®ã§èšç®ãããšããããèšç®ã®éçšã§âŠ ã®ããã«ãïŒâŠ
ãã€ãã³ã³ãã¯ãã«ããç¥èã玹ä»ããŠãã ãã ç¶ããïŒid:titiroboïŒããã®ããã°ã¯ãæ¯å楜ããèªãŸããŠããã ããŠããŸãã ä»åã¯ãã®èšäºã«ä¹ã£ããããŠããã ããŸãã www.nanigoto.net ãïŒã»ïŒã»ïŒã»ïŒã»ïŒããšäžŠãã æ°åããããšããŠããïŒãã®æ¬¡ã«âŠ
ãã€ãã®æªãçã§ããã£ãšæ©ãæžãããšæããªãããæ°ã¥ããããããããšåæ以äžçµã£ãŠããŸã£ãã Twitter ãžã®èšåãããã®ãšããå€ãããä»åã¯ã¿ã€ã ã©ã€ã³ã«ãããªãªãã€ãŒããæµããŠããããšããã£ããã ã£ããFFå€ããåŒçšå€±ç€ŒããŸãã ããã pic.twâŠ
ååã®ãšã³ããªãŒ ã®ç¶ãã§ããååã®ããããã¯ãä»®æ³é貚ã®äžææ²ç·ãææ°é¢æ°ã«ãŒãã«äŒŒãŠããã®ã§ãçç³»ã«ã¯ããªãã¿ã®ããžã¹ãã£ãã¯æ¹çšåŒ ããé 延æéãå å³ããŠãããªãµãã«æ¹é ãã Excel ã§æç·è¿äŒŒã«ããç°¡åãªã·ãã¥ã¬ãŒããè¡ã£ãããããã«ã®âŠ
ãªãã ãæ¥ã«ã²ããããã®ã§ãä»ãã£ãŠã話é¡ã¯ã²ãšãŸãæªããŠãå ã«ãã¡ããæžããŸãã ãããã§ä»®æ³é貚ã®è©±é¡ãçãäžãã£ãŠããŸããããã£ããã®äººãïŒããšããæå€ãªäººãŸã§ããæè³ã«åå ¥ããŠããããã§ãããããŸã§å人ã®ææ³ã§ãããŸãæ¬äººã«çŽæ¥ããâŠ
ãããŸã§ç§ã®èŠ³æž¬ç¯å²ã§ããããé»éæ¯ãéããªããŒã ã®ããã ãéããªããŒã ã£ãŠè¡æ£èãèšãæ¹ã ãèªåã§èšã£ãšããŠãªãã ãã çºç«¯ã¯ãã€ãªãŒããŒã¿ã«Zã®ãã®èšäºã ã£ããšæãã portal.nifty.com ãããåããŠãã©ããããé ã®æªã人ã®çµµããé»éæ¯ã«æ²¿âŠ
O'REILLYããŒãããäœãDeep Learning âPythonã§åŠã¶ãã£ãŒãã©ãŒãã³ã°ã®çè«ãšå®è£ ãïŒä»¥äž âããã¹ãâïŒèªè éå®å¯Ÿè±¡ã®ãäœåºŠç®ãã®ã·ãªãŒãºã®æçµåã§ãã ååã¯ã·ã°ã¢ã€ããš2ä¹å誀差ã®èª€å·®éäŒææ³ã®å®è£ ã«ã€ããŠæžããããããã¹ã5ç« ã®ã¯ã©ã€ããã¯âŠ
ååã®å 容ãããªãã¡ã2ç« ããŒã»ãããã³ã«ããè«çã²ãŒããã誀差éäŒææ³ã§æ©æ¢°åŠç¿ãããšããèªäœã®æŒç¿åé¡ã解ããŠã¿ãææ³ã ã ãŸãã¯ããªãèªåã®æžããã¹ã¯ãªããããåãã°ããããšããã¬ãã«ã«ãããæãããšããããšãç¹°ãè¿ããããšããçç±ãâŠ
5ç« ã§ã¯ãèªåèªèº«ã«åºé¡ããæŒç¿åé¡ã2åäœã£ãŠã¿ãã 1åç®ã¯ã6æ25æ¥ä»ãšã³ããªãŒã«æžããã2ç« ããŒã»ãããã³ã«ããè«çã²ãŒããã誀差éäŒææ³ã§æ©æ¢°åŠç¿ãããŠã¿ããããšãããã®ã ãåæ¥ä»ã§ã¯æ°å€åŸ®åæ³ã§è§£ããã watto.hatenablog.com ãŸãã¯ãâŠ
çžå€ãããO'REILLYããŒãããäœãDeep Learning âPythonã§åŠã¶ãã£ãŒãã©ãŒãã³ã°ã®çè«ãšå®è£ ãïŒä»¥äž âããã¹ãâïŒãå°ããã€èªãã§ãããèªãã°èªãã»ã©ãèè ã® æè€åº·æ¯ æ°ã¯ãã€ãã¥ãé ã®ãã人ãªãã ãšããããšãå®æãããã©ã®ç« ããç« ã®åãã«ãâŠ