ãæ©æ¢°åŠç¿ãããç¥çµå路網ããªã©ã®èšèã®æ§é ãããè³ã«ããŸãã ãããã®è¡šçŸã¯ãã§ã«äžè¬ã®äººã ã®æèã«å ¥ããã»ãšãã©ã®å Žåããã¿ãŒã³ãšé³å£°èªèã人éã®ãããªããã¹ãã®çæã«é¢é£ä»ããããŠããŸãã å®éãæ©æ¢°åŠç¿ã¢ã«ãŽãªãºã ã¯ãäžå°äŒæ¥ããªã³ã©ã€ã³åºçç©ãªã©ãããŸããŸãªçš®é¡ã®ã¿ã¹ã¯ã解決ã§ããŸãã ãã®èšäºã§ã¯ãã¹ã³ã¢ãªã³ã°ã¢ãã«ã®äœæãšããå®éã®ããžãã¹äžã®åé¡ã解決ã§ãããã¥ãŒã©ã«ãããã¯ãŒã¯ã®äœææ¹æ³ã説æããŸãã ããŒã¿ã®æºåããã¢ãã«ã®äœæãšãã®å質ã®è©äŸ¡ãŸã§ããã¹ãŠã®æ®µéãèæ ®ããŸãã
èšäºã§èª¬æãããŠãã質åïŒ
â¢ã¢ãã«ãæ§ç¯ããããã®ããŒã¿ãåéããã³æºåããæ¹æ³ã
â¢ãã¥ãŒã©ã«ãããã¯ãŒã¯ãšã¯äœã§ããã
â¢ãã¥ãŒã©ã«ãããã¯ãŒã¯ããŒãããèšè¿°ããæ¹æ³
â¢å©çšå¯èœãªããŒã¿ã§ãã¥ãŒã©ã«ãããã¯ãŒã¯ãé©åã«ãã¬ãŒãã³ã°ããæ¹æ³
â¢ã¢ãã«ãšãã®çµæã®è§£éæ¹æ³ã¯ïŒ
â¢ã¢ãã«ã®å質ãæ£ããè©äŸ¡ããæ¹æ³ã¯ïŒ
ãã³ã³ãã¥ãŒã¿ãŒãèããããšãã§ãããã©ãããšããåé¡ã¯ããã¯ãèå³æ·±ããã®ã§ã¯ãããŸããã
æœæ°ŽèŠãèªè¡ã§ãããã©ãããšãã質åããããã
ãšãã¬ãŒã»ãŠã£ãŒãã»ãã€ã¯ã¹ãã©
å€ãã®äŒæ¥ã§ã¯ãã»ãŒã«ã¹ãããŒãžã£ãŒãæœåšçãªé¡§å®¢ãšéä¿¡ãããã¢ãè¡ãã補åã«ã€ããŠè©±ããŸãã 圌ãã¯ãããã°ãããããå¶ç¶ã«æã«èœã¡ã人ã ã«ãã£ãŠã100åç®ã®éãã°ãã°ãã«åŒãè£ãããŸãã å€ãã®å Žåã顧客ã¯å¿ èŠãªãã®ããŸãã¯è£œåãæäŸã§ãããã®ãå®å šã«ç解ããŠããŸããã ãã®ãããªã¯ã©ã€ã¢ã³ããšã®ã³ãã¥ãã±ãŒã·ã§ã³ã¯ãåã³ãå©çããããããŸããã ãããŠãæãäžæå¿«ãªããšã¯ãæéå¶éã®ããã«ãæ¬åœã«éèŠãªã¯ã©ã€ã¢ã³ãã«ååãªæ³šæãæã£ãŠååŒãéãããšãã§ããªããšããããšã§ãã
ç§ã¯9æã®åæãµãŒãã¹Serpstatã®æ°åŠè ããã°ã©ããŒã§ãã æè¿ãæ¢åã®ã¹ã³ã¢ãªã³ã°ã¢ãã«ãšäœæ¥ã¹ã³ã¢ãªã³ã°ã¢ãã«ãæ¹åãã販売ã®æåã«åœ±é¿ããèŠå ãåè©äŸ¡ãããšããèå³æ·±ãã¿ã¹ã¯ãååŸããŸããã åŸç¹ã¯ã客æ§ã®èª¿æ»ã«åºã¥ããŠèæ ®ãããåé ç®ã¯è³ªåãžã®åçã«å¿ããŠãåèšåŸç¹ã«äžå®ã®ãã€ã³ããäžããŸããã ç°ãªã質åã«å¯Ÿããããããã¹ãŠã®ãã€ã³ãã¯ãçµ±èšç仮説ã«åºã¥ããŠé 眮ãããŸããã ã¹ã³ã¢ãªã³ã°ã¢ãã«ã䜿çšãããæéãçµéããããŒã¿ãåéãããããæ¥åœŒããç§ã®ãšããã«æ¥ãŸããã ããã§ååãªãµã³ãã«ãåŸãããã®ã§ãæ©æ¢°åŠç¿ã¢ã«ãŽãªãºã ã䜿çšããŠå®å šã«ä»®èª¬ãç«ãŠãããšãã§ããŸããã
ã¹ã³ã¢ãªã³ã°ã¢ãã«ã®äœææ¹æ³ã説æããŸãã ããã¯ãå®éã®ããžãã¹ã§çºçãããã¹ãŠã®å°é£ãšå¶éã䌎ããå®éã®ããŒã¿ã䜿çšããå®éã®ã±ãŒã¹ã§ãã ã ããããŸãæåã«ã
äœæ¥ã®ãã¹ãŠã®æ®µéã§è©³ãã説æããŸãã
âžããŒã¿åé
âžååŠç
modelã¢ãã«ã®æ§ç¯
âžå質åæãšã¢ãã«ã®è§£é
ããã€ã¹ããã¥ãŒã©ã«ãããã¯ãŒã¯ã®äœæãšãã¬ãŒãã³ã°ãæ€èšããŠãã ããã ãã®ãã¹ãŠã説æããå®éã®ã¹ã³ã¢ãªã³ã°ã®åé¡ã解決ããåžžã«æ°ããçè«ãäŸã§è£åŒ·ããŠããŸãã
ããŒã¿åé
æåã«ãå°æ¥ã®ã¢ãã«ã§ã¯ã©ã€ã¢ã³ãïŒãŸãã¯ãªããžã§ã¯ãïŒãè¡šãåé¡ãç解ããå¿ èŠããããŸãã ãããªãããã»ã¹ãããã«åºã¥ããŠæ§ç¯ãããŠãããããã¿ã¹ã¯ã«çå£ã«åãçµã¿ãŸãã ãŸãããªããžã§ã¯ãã説æããéèŠãªæ©èœãèŠéããŠã¯ãªããŸããã次ã«ãæ©èœã決å®ããããã®å³æ ŒãªåºæºãäœæããŸãã çµéšã«åºã¥ããŠã3ã€ã®ã«ããŽãªã®è³ªåãåºå¥ã§ããŸãã
- ããŒã«ïŒ2ã€ã®ã«ããŽãªïŒããã®çãã¯YesãŸãã¯NoïŒ1ãŸãã¯0ïŒã§ãã ããšãã°ã質åãžã®çãïŒã¯ã©ã€ã¢ã³ãã¯ã¢ã«ãŠã³ããæã£ãŠããŸããïŒ
- ã«ããŽãªãŒãç¹å®ã®ã¯ã©ã¹ã«å¯Ÿããçãã éåžžã3ã€ä»¥äžã®ã¯ã©ã¹ïŒãã«ãã«ããŽãªïŒããããŸãããã以å€ã®å Žåã¯ã質åãããŒã«ã«æžããããšãã§ããŸãã ããšãã°ãè²ïŒèµ€ãç·ãéã
- éçãç¹å®ã®ææšãç¹åŸŽä»ããæ°åã§ããçãã ããšãã°ãæãããã®ãããæ°ïŒ15ã
ãªããã®ããã«è©³çŽ°ã«ãã ããã®ã§ããïŒ éåžžãæ©æ¢°åŠç¿ã¢ã«ãŽãªãºã ã«ãã£ãŠè§£æ±ºãããå€å žçãªåé¡ãèãããšããæ°å€ããŒã¿ã®ã¿ãæ±ããŸãã ããšãã°ã20 x 20ãã¯ã»ã«ã®ç»åããã®çœé»ã®ææžãæ°åã®èªèã ãã®äŸã§ã¯ã400åã®æ°åïŒé»ãšçœã®ãã¯ã»ã«ã®æãããè¡šãïŒããµã³ãã«ã®1ã€ã®äŸãè¡šããŸãã äžè¬ã«ãããŒã¿ã¯æ°å€ã§ããå¿ èŠã¯ãããŸããã å®éãã¢ãã«ãæ§ç¯ããéã«ã¯ãã¢ã«ãŽãªãºã ãåŠçã§ãã質åã®çš®é¡ãç解ããå¿ èŠããããŸãã ããšãã°ããã¹ãŠã®çš®é¡ã®è³ªåã«ã€ããŠæ±ºå®æšãèšç·Žããããã¥ãŒã©ã«ãããã¯ãŒã¯ã¯æ°å€å ¥åããŒã¿ã®ã¿ãåãå ¥ããå®éçå±æ§ã«ã€ããŠã®ã¿èšç·ŽãããŸãã ããã¯ãããè¯ãã¢ãã«ã®ããã«ããã€ãã®åé¡ãæŸæ£ããå¿ èŠããããšããããšã§ããïŒ ããŒã¿ãæ£ããæºåããã ãã§ååã§ãã
ããŒã¿ã¯æ¬¡ã®å€å žçãªæ§é ãæã€å¿ èŠããããŸããåiçªç®ã®ã¯ã©ã€ã¢ã³ãã®ç¹åŸŽãã¯ãã«X ïŒiïŒ = {x ïŒiïŒ 1 ãx ïŒiïŒ 2 ã...ãx ïŒiïŒ n }ããã³ã¯ã©ã¹Y ïŒiïŒ -圌ããããè²·ã£ããã©ããã瀺ãã«ããŽãªã äŸïŒcustomer ïŒ3ïŒ = {greenãbitterã4.14ãyes}-è²·ããŸããã
åè¿°ã«åºã¥ããŠãããã«æºåããããã®è³ªåã®ã¿ã€ããå«ãããŒã¿åœ¢åŒãæ瀺ããããšããŸãã
ã¯ã©ã¹ïŒ
ïŒã«ããŽãªãŒïŒ | è²ïŒ
ïŒã«ããŽãªãŒïŒ | å³ïŒ
ïŒã«ããŽãªãŒïŒ | éãïŒ
ïŒæ°ïŒ | åºäœïŒ
ïŒããŒã«ïŒ |
---|---|---|---|---|
- | èµ€ | é žã£ã±ã | 4.23 | ã¯ã |
- | ç·è² | èŠã | 3.15 | ãã |
+ | ç·è² | èŠã | 4.14 | ã¯ã |
+ | é | çã | 4.38 | ãã |
- | ç·è² | å¡©èŸã | 3.62 | ãã |
è¡š1-ååŠçåã®ãã¬ãŒãã³ã°ãµã³ãã«ããŒã¿ã®äŸ
ååŠç
ããŒã¿ãåéããããããããæºåããå¿ èŠããããŸãã ãã®ã¹ãããã¯ååŠçãšåŒã°ããŸãã ååŠçã®äž»ãªã¿ã¹ã¯ã¯ãã¢ãã«ã®ãã¬ãŒãã³ã°ã«é©ãã圢åŒã§ããŒã¿ã衚瀺ããããšã§ãã ååŠç段éã§ã¯ã3ã€ã®äž»èŠãªããŒã¿æäœããããŸãã
- ãã¬ãŒãã³ã°ãµã³ãã«ã®äŸãååšããæšèã®ãã¯ãã«ç©ºéã®äœæã æ¬è³ªçã«ãããã¯ãã¹ãŠã®ããŒã¿ãæ°å€åœ¢åŒã«å€æããããã»ã¹ã§ãã ããã«ãããã«ããŽãªåãããŒã«åãããã³ãã®ä»ã®éæ°å€åããç§ãã¡ãæããŸãã
- ããŒã¿ã®æ£èŠåã ããšãã°ããã¹ãŠã®ããŒã¿ã«ãããåç¹åŸŽã®å¹³åå€ããŒãã§ãããåæ£ãåäœã§ãããšããããšãéæããããã»ã¹ã ããŒã¿ã®æ£èŠåã®æãå
žåçãªäŸã次ã«ç€ºããŸããX =ïŒX-ÎŒïŒ/Ï
æ£èŠåé¢æ°def normalize(X): return (XX.mean())/X.std()
- ãã¯ãã«ç©ºéã®æ¬¡å
ãå€æŽããŸãã ãã£ãŒãã£ã®ãã¯ãã«ç©ºéã倧ããããïŒæ°çŸäžã®ãã£ãŒãã£ïŒãŸãã¯å°ããããïŒ1ããŒã¹æªæºïŒå Žåãã¹ããŒã¹ã®æ¬¡å
ãå¢æžããæ¹æ³ãé©çšã§ããŸãã
- 次å ãå¢ããã«ã¯ãç¹åŸŽãã¯ãã«ã®ãããã®ãã€ã³ãã«è·é¢ãè¿œå ããããšã«ããããã¬ãŒãã³ã°ã»ããã®äžéšãåç §ãã€ã³ããšããŠäœ¿çšã§ããŸãã ãã®æ¹æ³ã¯ãé«æ¬¡å éåã®ç©ºéã§ã¯ç·åœ¢åé¢å¯èœã«ãªããšããäºå®ã«ã€ãªããããšãå€ããããã«ããåé¡åé¡ãç°¡çŽ åãããŸãã
- 次å ãäžããããã«ãPCAãæããã䜿çšãããŸãã äž»æåæ³ã®äž»ãªã¿ã¹ã¯ã¯ããã¬ãŒãã³ã°ãµã³ãã«ã®èŠçŽ ã®æ圱å€ã®åæ£ãæ倧åãããç¹åŸŽã®æ°ããç·åœ¢çµåãæ€çŽ¢ããããšã§ãã
ãã¯ãã«ç©ºéãæ§ç¯ããäžã§æãéèŠãªããªãã¯ã®1ã€ã¯ãã«ããŽãªåããã³ããŒã«åã®æ°ã®åœ¢åŒã§è¡šçŸããæ¹æ³ã§ãã äŒãïŒ ã¯ã³ããã ïŒãã·ã¢çµ±äžã³ãŒãïŒã ãã®ãããªãšã³ã³ãŒãã£ã³ã°ã®äž»ãªã¢ã€ãã¢ã¯ãã«ããŽãªå±æ§ããå¯èœãªã«ããŽãªã®æ°ã«å¯Ÿå¿ãã次å ãæã€ãã¯ãã«ç©ºéã®ãã¯ãã«ãšããŠæ瀺ããããšã§ãã ãã®å Žåããã®ã«ããŽãªã®åº§æšã®å€ã¯ãŠãããããšã«ååŸãããä»ã®ãã¹ãŠã®åº§æšã¯ãªã»ãããããŸãã ããŒã«å€ã䜿çšãããšããã¹ãŠãéåžžã«åçŽã«ãªããå®éã®åäœãŸãã¯ãŒãã«å€ãããŸãã
ããšãã°ããµã³ãã«é ç®ã¯ãèŠããçããå¡©èŸããé žã£ã±ãããŸãã¯ããŸå³ïŒèïŒã®ããããã§ãã ã¯ã³ããããšã³ã³ãŒãã¯æ¬¡ã®ããã«ãªããŸãïŒbitter =ïŒ1ã0ã0ã0ã0ïŒãsweet =ïŒ0ã1ã0ã0ã0ïŒãsalty =ïŒ0ã0ã1ã0ã0ïŒããµã¯ãŒ=ïŒ0ã0ã0ã0ã1ã0ïŒããã€ã³ã=ïŒ0ã0ã0ã0ã0ã1ïŒã ãªã4ã€ã®ãã¬ãŒããŒã§ã¯ãªã5ã€ã®ãã¬ãŒããŒãããã®ãââçåãããå Žåã¯ãå³èŠæèŠã·ã¹ãã ã«é¢ãããã®èšäºãèªãã§ãã ãã ãããããããã¯ã¹ã³ã¢ãªã³ã°ãšã¯é¢ä¿ãããŸããã4ã€ã䜿çšããŠãå€ãåé¡ã«éå®ããŸãã
ããã§ãã«ããŽãªãã£ãŒãã£ãéåžžã®æ°å€ãã¯ãã«ã«å€æããæ¹æ³ãåŠã³ãŸãããããã¯éåžžã«äŸ¿å©ã§ãã ããŒã¿ã®ãã¹ãŠã®æäœãå®äºããããã©ã®ã¢ãã«ã«ãé©ãããã¬ãŒãã³ã°ãµã³ãã«ãååŸããŸãã ç§ãã¡ã®å ŽåããŠãã¿ãªãšã³ã³ãŒãã£ã³ã°ãšæ£èŠåãé©çšãããšãããŒã¿ã¯æ¬¡ã®ããã«ãªããŸãã
ã¯ã©ã¹ïŒ | èµ€ïŒ | ç·ïŒ | éïŒ | èŠãïŒ | çãïŒ | ãµã«ãã£ïŒ | é žã£ã±ãïŒ | éãïŒ | åºäœïŒ |
---|---|---|---|---|---|---|---|---|---|
0 | 1 | 0 | 0 | 0 | 0 | 0 | 1 | 0.23 | 1 |
0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | -0.85 | 0 |
1 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0.14 | 1 |
1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0.38 | 0 |
0 | 0 | 1 | 0 | 0 | 0 | 1 | 0 | -0.48 | 0 |
è¡š2-ååŠçåŸã®ãã¬ãŒãã³ã°ãµã³ãã«ããŒã¿ã®äŸ
ååŠçãšã¯ã人éã«ãšã£ãŠã¯ããŸã䟿å©ã§ã¯ãªããæ©æ¢°ã奜ã圢åŒã§ç解ã§ããããŒã¿ã衚瀺ããããã»ã¹ã§ãããšèšããŸãã
ã»ãšãã©ã®å Žåãã¹ã³ã¢ãªã³ã°åŒã¯æ¬¡ã®ç·åœ¢ã¢ãã«ã§ãã
ããã§ãkã¯è³ªåè¡šã®è³ªåçªå·ãw kã¯åèšã¹ã³ã¢ãªã³ã°ã®ãã®kçªç®ã®è³ªåã«å¯Ÿããåçã®å¯äžä¿æ°ã§ã| w | -質åïŒãŸãã¯ä¿æ°ïŒã®æ°ãx k-ãã®è³ªåã«å¯Ÿããçãã ããã«ã質åã¯ãããŒã«ïŒyesãŸãã¯noã1ãŸãã¯0ïŒãæ°å€ïŒããšãã°ã身é·= 175ïŒããŸãã¯ã«ããŽãªã«ã«ã§ããããŠãã¿ãªãŒãšã³ã³ãŒãã£ã³ã°ã®åœ¢åŒïŒãªã¹ãããç·ïŒèµ€ãç·ããŸãã¯éïŒã§èšè¿°ã§ããŸãã = [0ã1ã0]ïŒã ããã«ãã«ããŽãªåã®è³ªåã¯ãåçãªãã·ã§ã³ã®ã«ããŽãªãšåãæ°ã®ããŒã«åã®è³ªåã«åé¡ããããšä»®å®ã§ããŸãïŒããšãã°ãã¯ã©ã€ã¢ã³ãã®èµ€ïŒã¯ã©ã€ã¢ã³ãã®ç·ïŒã¯ã©ã€ã¢ã³ãã®éïŒïŒã
ã¢ãã«éžæ
ä»æãéèŠãªããšïŒã¢ãã«ã®éžæã çŸåšãã¹ã³ã¢ãªã³ã°ã¢ãã«ãæ§ç¯ã§ããæ©æ¢°åŠç¿ã¢ã«ãŽãªãºã ã«ã¯ããã£ã·ãžã§ã³ããªãŒïŒæ±ºå®æšïŒãKNNïŒkæè¿åæ³ïŒãSVMïŒãµããŒããã¯ãã«æ³ïŒãNNïŒãã¥ãŒã©ã«ãããã¯ãŒã¯ïŒããããŸãã ãããŠãã¢ãã«ã®éžæã¯ãç§ãã¡ãããããæããã®ã«åºã¥ããŠããã¹ãã§ãã ãŸããã¢ãã«ã®çµæã«åœ±é¿ãäžãã決å®ãã©ãã ãæ確ã§ããå¿ èŠããããŸãã ã€ãŸããã¢ãã«ã®æ§é ã解éã§ããããšãã©ãã»ã©éèŠããšããããšã§ãã
å³ 1-æ©æ¢°åŠç¿ã¢ã«ãŽãªãºã ã®æè»æ§ãšçµæã®ã¢ãã«ã®è§£éå¯èœæ§ã®äŸåæ§
ããã«ããã¹ãŠã®ã¢ãã«ãç°¡åã«æ§ç¯ã§ããããã§ã¯ãªããäžéšã®ã¢ãã«ã«ã¯éåžžã«å ·äœçãªã¹ãã«ãšéåžžã«åŒ·åãªããŒããŠã§ã¢ãå¿ èŠã§ãã ããããæãéèŠãªããšã¯ãæ§ç¯ãããã¢ãã«ã®å®è£ ã§ãã ããžãã¹ããã»ã¹ãæ¢ã«ç¢ºç«ãããŠãããè€éãªã¢ãã«ãå°å ¥ããããšã¯ãŸã£ããäžå¯èœã§ãã ãŸãã¯ãã¯ã©ã€ã¢ã³ãã質åã«åçãããšãã«ãåçãªãã·ã§ã³ã«å¿ããŠãã©ã¹ãŸãã¯ãã€ãã¹ã®ãã€ã³ããåãåãç·åœ¢ã¢ãã«ãå¿ èŠã§ãã å Žåã«ãã£ãŠã¯ãéã«å®è£ ã®å¯èœæ§ããããå ¥åãã©ã¡ãŒã¿ãŒã®éåžžã«æçœã§ãªãçµã¿åãããèæ ®ã«å ¥ããè€éãªã¢ãã«ã§ããããããã®éã®çžäºæ¥ç¶ãèŠã€ããŸãã ããã§ãäœãéžã¶ã¹ãã§ããïŒ
æ©æ¢°åŠç¿ã¢ã«ãŽãªãºã ãéžæããéã«ããã¥ãŒã©ã«ãããã¯ãŒã¯ã«æ±ºããŸããã ãªãã§ïŒ ãŸããTensorFlowãTheanoãªã©ã®å€ãã®ã¯ãŒã«ãªãã¬ãŒã ã¯ãŒã¯ããããŸãã ãããã¯ãã¢ãŒããã¯ãã£ãšåŠç¿ãã©ã¡ãŒã¿ãéåžžã«æ·±ãçå£ã«ã«ã¹ã¿ãã€ãºããæ©äŒãæäŸããŸãã 第äºã«ãã¢ãã«ããã€ã¹ããå¶ç¶ã«ãååã«è§£éãããåå±€ãã¥ãŒã©ã«ãããã¯ãŒã¯ãããæ°è¡ã®ã³ãŒããå€æŽããã ãã§ãéç·åœ¢ã®äŸåé¢ä¿ãèŠã€ããåªããèœåãæã€å€å±€ãããã¯ãŒã¯ã«å€æŽããæ©èœã ããã«ãèšç·Žãããåå±€ãã¥ãŒã©ã«ãããã¯ãŒã¯ã¯ãããŸããŸãªã¢ã³ã±ãŒãã®è³ªåãžã®åçã®ãã€ã³ããå ç®ããå€å žçãªå æ³ã¹ã³ã¢ãªã³ã°ã¢ãã«ã«å€ããããšãã§ããŸãããããã«ã€ããŠã¯åŸã§è©³ãã説æããŸãã
ä»å°ãçè«ã ãã¥ãŒãã³ã掻æ§åé¢æ°ãæ倱é¢æ°ãåŸé éäžããšã©ãŒããã¯äŒææ³ãªã©ããã€ãã£ãã¯ãŒãã§ããå Žåã¯ããã®ãã¹ãŠãå®å šã«ã¹ãããã§ããŸãã ããã§ãªãå Žåã¯ãçæéã®äººå·¥ãã¥ãŒã©ã«ãããã¯ãŒã¯ãžããããã
人工ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®çæã³ãŒã¹
ãŸãã人工ãã¥ãŒã©ã«ãããã¯ãŒã¯ïŒANNïŒã¯ãå®éã®çç©åŠçãã¥ãŒã©ã«ãããã¯ãŒã¯ïŒANNïŒã®çµç¹ã®æ°åŠã¢ãã«ã§ãã ããããBNSã®æ°åŠã¢ãã«ãšã¯ç°ãªããANNã¯ã掻åé»äœïŒPDïŒã®ãççŒãã®èª¬æãç¥çµäŒéç©è³ªãã€ãªã³ãã£ãã«ãäºæ¬¡äžéäœã茞éã¿ã³ãã¯è³ªãªã©ã®ãã¹ãŠã®ååŠçããã³ç©ççããã»ã¹ã®æ£ç¢ºãªèª¬æãå¿ èŠãšããŸãããå®éã®BNSã®äœæ¥ã¯ãç©çã¬ãã«ã§ã¯ãªããæ©èœã¬ãã«ã§ã®ã¿ã§ãã
ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®åºæ¬èŠçŽ ã¯ãã¥ãŒãã³ã§ãã ãã¥ãŒãã³ã®æãåçŽãªæ©èœçãªæ°åŠã¢ãã«ãäœæããŠã¿ãŸãããã ãããè¡ãããã«ãäžè¬çãªçšèªã§çç©åŠçãã¥ãŒãã³ã®æ©èœã説æããŸãã
å³ 2-çäœãã¥ãŒãã³ã®å žåçãªæ§é
ã芧ã®ãšãããçäœãã¥ãŒãã³ã®æ§é ã¯ãæš¹ç¶çªèµ·ããã¥ãŒãã³ã®æ¬äœãããã³è»žçŽ¢ã«åçŽåã§ããŸãã æš¹ç¶çªèµ·ã¯ããã¥ãŒãã³ãžã®å ¥ãå£ããæ å ±ãåéããåå²ããã»ã¹ã§ãïŒããã¯ãå容äœããã®å€éšæ å ±ãããšãã°è²ã®å Žåã¯éäœããããŸãã¯å¥ã®ãã¥ãŒãã³ããã®å éšæ å ±ïŒã çä¿¡æ å ±ããã¥ãŒãã³ã掻æ§åããå ŽåïŒçç©åŠçãªå Žåãé»äœãç¹å®ã®ãããå€ãããé«ããªã£ãïŒãå±èµ·æ³¢ïŒAPïŒãçæããããã¥ãŒãã³ã®äœã®èãäŒæããç¥çµäŒéç©è³ªã®æŸåºãä»ããŠè»žçŽ¢ãä»ããŠä»ã®ç¥çµçŽ°èã«ä¿¡å·ãéä¿¡ããŸã现èãŸãã¯çµç¹ã
ããã«åºã¥ããŠããŠã©ãŒã¬ã³ããã«ããã¯ãšãŠã©ã«ã¿ãŒãããã¯1943幎ã«æ°åŠçãã¥ãŒãã³ã¢ãã«ãææ¡ããŸããã ãããŠã1958幎ãããã«ããã¯ããããã¥ãŒãã³ã«åºã¥ãããã©ã³ã¯ããŒãŒã³ãã©ãããã³ã³ãã¥ãŒã¿ãŒããã°ã©ã ãäœæãã次ã«ç©çããã€ã¹-ããŒã»ãããã³ãäœæããŸããã ããããã人工ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®æŽå²ãå§ãŸããŸããã 次ã«ããã¥ãŒãã³ã®æ§é ã¢ãã«ã«ã€ããŠèããŠã¿ãŸããããããã«ã€ããŠã¯ãããã«å¯ŸåŠããŸãã
å³ 3-McCallock-Pittsæ°åŠãã¥ãŒãã³ã¢ãã«
ã©ãã§ïŒ
- Xã¯å
¥åãã©ã¡ãŒã¿ãŒãã¯ãã«ã§ãã ãã¥ãŒãã³ã®å
¥åã«ãªã£ãæ°åã®ãã¯ãã«ïŒåïŒïŒbiolãããŸããŸãªå容äœã®æŽ»æ§åã®çšåºŠïŒã
Wã¯ãéã¿ã®ãã¯ãã«ïŒäžè¬çãªå Žåãéã¿ã®è¡åïŒãåŠç¿ããã»ã¹äžã«å€åããæ°å€ïŒbiolãã·ããã¹å¯å¡æ§ã«åºã¥ããã¬ãŒãã³ã°ããã¥ãŒãã³ã¯ãã®å容äœããã®ä¿¡å·ã«æ£ããå¿çããããšãåŠç¿ããŸãïŒã - å ç®åšã¯ã察å¿ããéã¿ã§ä¹ç®ããããã¹ãŠã®å ¥åãã©ã¡ãŒã¿ãŒãå ç®ãããã¥ãŒãã³ã®æ©èœãããã¯ã§ãã
- ãã¥ãŒãã³ã®æŽ»æ§åã®æ©èœã¯ãå ç®åšããã®å€ã«å¯Ÿãããã¥ãŒãã³ã®åºåã®å€ã®äŸåé¢ä¿ã§ãã
- 次ã®ãã¥ãŒãã³ãç¹å®ã®ãã¥ãŒãã³ã®åºåããã®å€ããç¬èªã®å ¥åã»ããã®1ã€ã«äŸçµŠãããŸãïŒãã®ãã¥ãŒãã³ãæåŸã®ã¿ãŒããã«ã§ããå Žåããã®å±€ã¯ååšããªãå ŽåããããŸãïŒã
æ°åŠçãã¥ãŒãã³ã®å®è£
import numpy as np def neuron(x, w): z = np.dot(w, x) output = activation(z) return output
次ã«ããããã®æå°æ§é åäœãããå€å žçãªäººå·¥ãã¥ãŒã©ã«ãããã¯ãŒã¯ãçµã¿ç«ãŠãããŸãã 次ã®çšèªãæ¡çšãããŸããã
- å ¥åïŒå容äœïŒå±€ã¯ããã©ã¡ãŒã¿ãŒïŒå±æ§ïŒã®ãã¯ãã«ã§ãã ãã®å±€ã¯ãã¥ãŒãã³ã§æ§æãããŠããŸããã ããã¯ããå€éšãã®äžçããã®å容äœã«ãã£ãŠååŸãããããžã¿ã«æ å ±ã§ãããšèšããŸãã ç§ãã¡ã®å Žåãããã¯é¡§å®¢æ å ±ã§ãã ã¬ã€ã€ãŒã«ã¯ãå ¥åãã©ã¡ãŒã¿ãŒãšåæ°ã®èŠçŽ ãå«ãŸããŸãïŒããã«ãã¢ã¯ãã£ãåãããå€ãã·ããããããã«å¿ èŠãªãã€ã¢ã¹é ãå«ãŸããŸãïŒã
- é£æ³ïŒé衚瀺ïŒã¬ã€ã€ãŒã¯ãäŸãèšæ¶ããè€éãªçžé¢é¢ä¿ãšéç·åœ¢äŸåé¢ä¿ãèŠã€ããæœè±¡åãšäžè¬åãæ§ç¯ã§ããæ·±ãæ§é ã§ãã äžè¬çãªã±ãŒã¹ã§ã¯ãããã¯ã¬ã€ã€ãŒã§ã¯ãªããå ¥åãšåºåã®éã®å€æ°ã®ã¬ã€ã€ãŒã§ãã åã¬ã€ã€ãŒã¯ã次ã®ã¬ã€ã€ãŒã®ããã«æ°ããïŒããé«ãã¬ãã«ã®ïŒç¹åŸŽãã¯ãã«ãæºåããŠãããšèšããŸãã é«ã¬ãã«ã®æœè±¡åã®åŠç¿ããã»ã¹ã«ãããå€èŠ³ã®åå ã¯ããã®å±€ã§ãã æ§é ã«ã¯ãå¿ èŠãªæ°ã®ãã¥ãŒãã³ãšå±€ãå«ãŸããååšããªãå ŽåããããŸãïŒç·åœ¢åé¢å¯èœã»ããã®åé¡ã®å ŽåïŒã
- åºåå±€ã¯ãåãã¥ãŒãã³ãç¹å®ã®ã¯ã©ã¹ãæ åœããå±€ã§ãã ãã®ã¬ã€ã€ãŒã®åºåã¯ãç°ãªãã¯ã©ã¹ã«å±ãããªããžã§ã¯ãã®ç¢ºçååžã®é¢æ°ãšããŠè§£éã§ããŸãã ã¬ã€ã€ãŒã«ã¯ããã¬ãŒãã³ã°ã»ããã«ããã¯ã©ã¹ãšåãæ°ã®ãã¥ãŒãã³ãå«ãŸããŸãã 2ã€ã®ã¯ã©ã¹ãããå Žåã2ã€ã®åºåãã¥ãŒãã³ã䜿çšãããã1ã€ã ãã«å¶éããããšãã§ããŸãã ãã®å Žåã1ã€ã®ãã¥ãŒãã³ã¯äŸç¶ãšããŠ1ã€ã®ã¯ã©ã¹ã®ã¿ãæ åœããŸããããŒãã«è¿ãå€ãçæããå Žåããã®ããžãã¯ã«ãããµã³ããªã³ã°èŠçŽ ã¯å¥ã®ã¯ã©ã¹ã«å±ããŠããå¿ èŠããããŸãã
å³ 4-å ¥åïŒå容äœïŒãåºåãã¯ã©ã¹ææ決å®è ãããã³é£æ³ïŒé衚瀺ïŒå±€ãåããå€å žçãªãã¥ãŒã©ã«ãããã¯ãŒã¯ããããž
é ãããé£æ³å±€ã®ååšã«ããã人工ãã¥ãŒã©ã«ãããã¯ãŒã¯ã¯è€éãªäŸåé¢ä¿ã®çºèŠã«åºã¥ããŠä»®èª¬ãç«ãŠãããšãã§ããŸãã ããšãã°ãç»åãèªèããç³ã¿èŸŒã¿ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®å Žåãç»åãã¯ã»ã«ã®èŒåºŠã¯å ¥åå±€ã«éä¿¡ãããåºåå±€ã«ã¯ç¹å®ã®ã¯ã©ã¹ïŒäººéãæ©æ¢°ãæš¹æšã家ãªã©ïŒãæ åœãããã¥ãŒãã³ãå«ãŸããŸããé ãå±€ã®ãå容äœãã¯ãããŸããŸãªåŸæè§ã®çŽç·ããè奮ããè§åºŠãæ£æ¹åœ¢ãåãåå§ãã¿ãŒã³ã«åå¿ãããã¥ãŒãã³ããç¹åãïŒç¹åïŒãå§ããŸãã ntã åºåå±€ã®è¿ãã«ã¯ãããšãã°ç®ãè»èŒªã錻ãç¿Œãèãé¡ãªã©ã«åå¿ãããã¥ãŒãã³ããããŸãã
å³ 5-ç³ã¿èŸŒã¿ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®åŠç¿ããã»ã¹ã«ãããéå±€çãªé¢é£ä»ãã®åœ¢æ
çç©åŠçã¢ãããžãŒãåŒãåºããŠãã¹ããŒãã¢ãã«ã«é¢ãã泚ç®ãã¹ãç¥çµççåŠè Vyacheslav Albertovich Dubyninã®èšèãåç §ããããšæããŸãã
ãç§ãã¡ã®è³ã¯ãäžäœã¬ãã«ã®åèªãèŠçŽããåèªãäœæãçæã§ããŸãã ãããŒãããŒã«ããã¥ãŒãã人圢-ããã¡ãã ããã¡ããæãå®¶å ·ã¯ãªããžã§ã¯ãã§ãã ãªããžã§ã¯ãã家ã人ã¯ç°å¢ã®ãªããžã§ã¯ãã§ãã ãããŠããå°ããç§ãã¡ã¯æœè±¡çãªå²åŠçæŠå¿µãæ°åŠçãç©ççæŠå¿µã«å°éããŸãã ã€ãŸããé³å£°ã®äžè¬åã¯ãé£åæ§é é ç®è³ªã®éåžžã«éèŠãªç¹æ§ã§ãããããã«ããã«ãã¬ãã«ã§ãããå€éšäžçå šäœã®é³å£°ã¢ãã«ã圢äœãããšãã§ããŸãã ããæç¹ã§ãç¥çµã€ã³ãã«ã¹ã¯ãã®é³å£°ã¢ãã«ã«æ²¿ã£ãŠéåžžã«æŽ»çºã«åãããšãã§ãããã®åããèªãã«æãèšèãæèããšåŒã³ãŸãã
ããããã®çè«ïŒïŒ ããããè¯ããã¥ãŒã¹ããããŸããæãåçŽãªå Žåããã¥ãŒã©ã«ãããã¯ãŒã¯å šäœãåäžã®ãã¥ãŒãã³ã§è¡šãããšãã§ããŸãã ããã«ãç¹ã«ãããã®ã¯ã©ã¹ã®ãªããžã§ã¯ããç·åœ¢ã«åé¢å¯èœãªç©ºéå ã®ãªããžã§ã¯ãã®ã¯ã©ã¹ãèªèããããšã«ãªããšã1ã€ã®ãã¥ãŒãã³ã§ããã¿ã¹ã¯ã«å¯ŸåŠããããšããããããŸãã å€ãã®å Žåãç·åœ¢ã®åé¢å¯èœæ§ã¯ãäžèšã®ããã«ç©ºéã®æ¬¡å ã倧ããããããšã§å®çŸã§ãã1ã€ã®ãã¥ãŒãã³ã«éå®ãããŸãã ãã ãããã¥ãŒã©ã«ãããã¯ãŒã¯ã«ããã€ãã®é ãå±€ãè¿œå ããæ¹ãç°¡åãªå Žåãããããµã³ãã«ããã®ç·åœ¢åé¢æ§ã¯å¿ èŠãããŸããã
å³ 6-ç·åœ¢ã«åé¢å¯èœãªã»ãããšç·åœ¢ã«åé¢ã§ããªãã»ãã
ããã§ã¯ããã®ãã¹ãŠãæ£åŒã«èª¬æããŸãããã ãã¥ãŒãã³ã®å ¥åã«ã¯ããã©ã¡ãŒã¿ãŒã®ãã¯ãã«ããããŸãã ç§ãã¡ã®å Žåããããã¯æ°å€åœ¢åŒX ïŒiïŒ = {x ïŒiïŒ 1 ãx ïŒiïŒ 2 ã...ãx ïŒiïŒ n }ã§è¡šãããã¯ã©ã€ã¢ã³ã調æ»ã®çµæã§ãã ããã«ãåã¯ã©ã€ã¢ã³ãã¯Y ïŒiïŒ -ãªãŒãã®æåãç¹åŸŽä»ããã¯ã©ã¹ïŒ1ãŸãã¯0ïŒã«é¢é£ä»ããããŠããŸãã å®éããã¥ãŒã©ã«ãããã¯ãŒã¯ã¯ã次å ãç¹åŸŽã®æ°ã«å¯Ÿå¿ãããã¯ãã«ç©ºéã§æé©ãªåé¢è¶ æ²é¢ãèŠã€ããªããã°ãªããŸããã ãã®å Žåããã¥ãŒã©ã«ãããã¯ãŒã¯ã®ãã¬ãŒãã³ã°ã¯ãã¯ã©ã¹ãæ åœãããã¥ãŒãã³ã顧客ãè³Œå ¥ããå Žåã«1ã«è¿ãå€ãçæããããã§ãªãå Žåã¯ãŒãã«è¿ãå€ãçæããéã¿è¡åWã®å€ïŒä¿æ°ïŒãèŠã€ããããšã§ãã
åŒãããããããã«ããã¥ãŒãã³ã®ä»äºã®çµæã¯ãå ¥åãã©ã¡ãŒã¿ãŒãšåŠç¿ããã»ã¹ã§æ±ããããä¿æ°ã®ç©ã®åèšããã®æŽ»æ§åé¢æ°ïŒå€ãã®å Žåã§ç€ºãããïŒã§ãã ã¢ã¯ãã£ããŒã·ã§ã³æ©èœãšã¯äœããèŠãŠã¿ãŸãããã
ä»»æã®å®æ°å€ããã¥ãŒã©ã«ãããã¯ãŒã¯ã®å ¥åã«ãªããéã¿è¡åã®ä¿æ°ãä»»æã«ãªãå¯èœæ§ãããããããããã®ç©ã®åèšã®çµæã¯è² ããæ£ã®ç¡é倧ãŸã§ã®ä»»æã®å®æ°ã«ãªããŸãã ãã¬ãŒãã³ã°ã»ããã®åèŠçŽ ã«ã¯ããã®ãã¥ãŒãã³ã«é¢é£ããã¯ã©ã¹å€ïŒ0ãŸãã¯1ïŒããããŸãã 0ãã1ãŸã§ã®åãç¯å²ã®ãã¥ãŒãã³ããå€ãååŸãããã®å€ãè¿ãå Žæã«å¿ããŠã¯ã©ã¹ã決å®ããããšããå§ãããŸãã ãã®å€ããèŠçŽ ããã®ã¯ã©ã¹ã«å±ãã確çãšããŠè§£éããã®ãããã«è¯ãã§ãã ãããã£ãŠãå®æ°ã®ã»ããã®èŠçŽ ã0ãã1ã®é åã«ãããã³ã°ãããããªå調ã§æ»ãããªé¢æ°ãå¿ èŠã§ãã ããããã·ã°ã¢ã€ãã¯ãã®äœçœ®ã«æé©ã§ãã
å³ 7-ã·ã°ã¢ã€ãã¯ã©ã¹ã®æãå€å žçãªä»£è¡šã®1ã€ã§ããããžã¹ãã£ãã¯ã«ãŒãã®ã°ã©ã
ã¢ã¯ãã£ããŒã·ã§ã³æ©èœ
def activation(z): return 1/(1+np.exp(-z))
ãšããã§ãå®éã®çäœãã¥ãŒãã³ã§ã¯ããã®ãããªé£ç¶çãªæŽ»æ§åæ©èœã¯å®çŸãããŠããŸããã ç§ãã¡ã®çŽ°èã«ã¯ãå¹³å-70mVã®éæ¢é»äœããããŸãã ãã¥ãŒãã³ã«æ å ±ãäŸçµŠããããšã掻æ§åãããå容äœã¯ãããšçµåããã€ãªã³ãã£ãã«ãéãã现èå ã®é»äœã®å¢å ãŸãã¯æžå°ã«ã€ãªãããŸãã å容äœã®æŽ»æ§åã«ãããååãšããã¬ãŒãã³ã°äžã«åŸãããéã¿è¡åã®1ã€ã®ä¿æ°ãšã®éã«é¡äŒŒæ§ãæãããšãã§ããŸãã é»äœã-50mVã®å€ã«éãããšããã«PDãçºçããè奮波ã¯è»žçŽ¢ã«æ²¿ã£ãŠã·ããã¹å端ã«å°éããç¥çµäŒéç©è³ªãã·ããã¹éç°å¢ã«æã蟌ã¿ãŸãã ã€ãŸããå®éã®çç©åŠç掻æ§åã¯æ»ããã§ã¯ãªã段éçã§ãããã¥ãŒãã³ã¯æŽ»æ§åãããããããŸããã ããã¯ãã¢ãã«ã®æ§ç¯ã«ãããŠæ°åŠçã«èªç±ã§ããããšã瀺ããŠããŸãã èªç¶ããåæ£ã³ã³ãã¥ãŒãã£ã³ã°ãšåŠç¿ã®åºæ¬åçãåãå ¥ããŠãä»»æã®ç¹æ§ãæã€èŠçŽ ã§æ§æãããèšç®ã°ã©ããæ§ç¯ã§ããŸãã ãã®äŸã§ã¯ããã¥ãŒãã³ããé¢æ£å€ã§ã¯ãªãé£ç¶å€ãåãåããŸãã äžè¬çãªå Žåãã¢ã¯ãã£ããŒã·ã§ã³æ©èœã¯ç°ãªãå ŽåããããŸãã
äžèšããåŠç¿ããæãéèŠãªããšã¯æ¬¡ã®ãšããã§ãã ããã¥ãŒã©ã«ãããã¯ãŒã¯ãã¬ãŒãã³ã°ïŒã·ããã¹ãã¬ãŒãã³ã°ïŒã¯ããšã©ãŒãæå°åããããã«éã¿è¡åã®ä¿æ°ã®æé©ãªéžæã«æžããå¿ èŠããããŸãããã¯ã©ã¹ã
ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®äœæ¥ã®çµæã¯ãéåžžã仮説ïŒè±èªã®ä»®èª¬ïŒãšåŒã°ããŸãã hïŒXïŒã§è¡šãããªããžã§ã¯ãã®å ¥åãã£ãŒãã£ïŒãã©ã¡ãŒã¿ãŒïŒã«å¯Ÿãã仮説ã®äŸåæ§ã瀺ããŸãã ãªã仮説ãªã®ãïŒ ããã¯æŽå²çã«äºå®ã§ãã å人çã«ã¯ãç§ã¯ãã®çšèªã奜ãã§ãã ãã®çµæããã¥ãŒã©ã«ãããã¯ãŒã¯ã®ä»®èª¬ãçŸå®ïŒãªããžã§ã¯ãã®å®éã®ã¯ã©ã¹ïŒã«å¯èœãªéã察å¿ãããããšèããŠããŸãã å®éãããã§ã¯çµéšããåŠã¶ãšããäž»èŠãªã¢ã€ãã¢ãçãŸããŸãã ããã§ããã¥ãŒã©ã«ãããã¯ãŒã¯ã®å質ãè¡šã尺床ãå¿ èŠã§ãã ãã®æ©èœã¯ãéåžžãæ倱æ©èœããšåŒã°ããŸãã éåžžãæ©èœã¯JïŒWïŒã§ç€ºãããéã¿è¡åã®ä¿æ°ãžã®äŸåæ§ã瀺ããŸãã æ©èœæ§ãäœãã»ã©ããã¥ãŒã©ã«ãããã¯ãŒã¯ããã¹ãç¯ãå¯èœæ§ã¯äœããªããŸãã ãã®æ©èœãæå°éã«æããããã«ãã¬ãŒãã³ã°ãè¡ãããŸãã éã¿è¡åã®ä¿æ°ã«å¿ããŠããã¥ãŒã©ã«ãããã¯ãŒã¯ã®ç²ŸåºŠãç°ãªãå ŽåããããŸãã åŠç¿ããã»ã¹ã¯æ倱æ±é¢æ°ã®è¶ æ²é¢ã«æ²¿ã£ãåãã§ããããã®ç®çã¯ãã®é¢æ°ãæå°åããããšã§ãã
å³ 8-æ倱é¢æ°ã®æ¥µå°å€ãžã®åŸé éäžãšããŠã®åŠç¿ããã»ã¹
éåžžãéã¿è¡åã®ä¿æ°ã¯ã©ã³ãã ã«åæåãããŸãã åŠç¿ããã»ã¹ã§ã¯ãä¿æ°ãå€åããŸãã ã°ã©ãã¯ãè¿åã§åæåããããã¥ãŒã©ã«ãããã¯ãŒã¯ã®éã¿è¡åã®ä¿æ°w 1ããã³w 2ã®å€åãšããŠã2ã€ã®ç°ãªãå埩åŠç¿ãã¹ã瀺ããŠããŸãã
次ã«ããã¥ãŒã©ã«ãããã¯ãŒã¯ã®ãã¬ãŒãã³ã°æ¹æ³ã«ã€ããŠèª¬æããŸããããã«ã¯å€ãã®ãªãã·ã§ã³ããããŸããã2ã€ã«ã€ããŠã話ããŸããé²åçïŒéºäŒçïŒã¢ã«ãŽãªãºã ãšåŸé éäžæ³ã§ãããããã®æ¹æ³ã®äž¡æ¹ã䜿çšãããŸããé²åã¢ã«ãŽãªãºã ã¯ãèªç¶selectionã®ã¢ããªã³ã°ã«åºã¥ãã人工ç¥èœã®æ¹åã§ããé²åçæææ³ã¯éåžžã«ç解ãããããåå¿è ã«ãšã£ãŠã¯ããè¯ãæ¹æ³ã§ããçŸåšã§ã¯ãäž»ã«ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®æ·±å±€ã®ãã¬ãŒãã³ã°ã«äœ¿çšãããŠããŸããåŸé éäžæ³ãšèª€å·®ã®éäŒææ³ã¯ããè€éã§ãããæãå¹æçã§äžè¬çãªæè²æ¹æ³ã®1ã€ã§ãã
é²åçåŠç¿
ãã®æ¹æ³ã®æ çµã¿ã§ã¯ã次ã®çšèªã䜿çšããŸãïŒéã¿è¡åã®ä¿æ°-ã²ãã ã1ã€ã®ä¿æ°-éºäŒåãæ倱ã®ãéããŸã®æ倱ãé¢æ°-ãã£ãããã¹ã©ã³ãã¹ã±ãŒãïŒããã§ã¯ããã§ã«å±ææ倧å€ãæ¢ããŠããŸãããããã¯åãªãæ £ç¿ã§ãïŒããã®æ¹æ³ã¯æ¬åœã«ç°¡åã§ãããã¥ãŒã©ã«ãããã¯ãŒã¯ã®ããããžïŒããã€ã¹ïŒãéžæãããã次ã®ããšãè¡ãå¿ èŠããããŸãã
- -1ã1ã®ç¯å²ã§ã©ã³ãã ã«ã²ãã ïŒéã¿è¡åïŒãåæåããŸãããããæ°åç¹°ãè¿ããŠãç°ãªããã©ã³ãã ãªãã¥ãŒã©ã«ãããã¯ãŒã¯ã®åæéå£ãäœæããŸããæ¯éå£ã®ãµã€ãºã¯ãP-æ¯éå£ãŸãã¯èŠªã«ãã£ãŠç€ºãããŸãã
éã¿è¡åã®ä¿æ°ã®ã©ã³ãã åæåimport random def generate_population(p, w_size): population = [] for i in range(p): model = [] for j in range(w_size + 1): # +1 for b (bias term) model.append(2 * random.random() - 1) # random initialization from -1 to 1 for b and w population.append(model) return np.array(population)
- . , - , () . : , -0.1 0.1.
def mutation(genom, t=0.5, m=0.1): mutant = [] for gen in genom: if random.random() <= t: gen += m*(2*random.random() -1) mutant.append(gen) return mutant
- , , ( â , â ). .
def accuracy(X, Y, model): A = 0 m = len(Y) for i, y in enumerate(Y): A += (1/m)*(y*(1 if neuron(X[i], model) >= 0.5 else 0)+(1-y)*(0 if neuron(X[i], model) >= 0.5 else 1)) return A
- « » P . 2, . : 80%.
def selection(offspring, population): offspring.sort() population = [kid[1] for kid in offspring[:len(population)]] return population
def evolution(population, X_in, Y, number_of_generations, children): for i in range(number_of_generations): X = [[1]+[v.tolist()] for v in X_in] offspring = [] for genom in population: for j in range(children): child = mutation(genom) child_loss = 1 - accuracy(X_in, Y, child) # or child_loss = binary_crossentropy(X, Y, child) is better offspring.append([child_loss, child]) population = selection(offspring, population) return population
ãã®ãããªç¥çµé²åã¯æ¹åããããšãã§ããŸããããšãã°ãÏã¯çªç¶å€ç°èªçºçãÎŒã¯çªç¶å€ç°èªçºã®åŒ·ããªã©ãè¿œå ã®ãã©ã¡ãŒã¿ãŒéºäŒåãå°å ¥ã§ããŸããããã§ããã¥ãŒãã³ã®éã¿ã®è¡åã«çžå çãªçªç¶å€ç°ã確çÏã§å°å ¥ãããéžæããç¯å²ïŒããšãã°ã-0.1ãã0.1ïŒã®ÎŒã®åãã©ã¡ãŒã¿ãŒã«ä¹±æ°ãè¿œå ãããŸãããããã®éºäŒåãå€åããããã§ãããã
çæ³çã«ã¯ãéžæã¯é²åã®ããŸããŸãªæ®µéã§å€ç°èªçºã®åŒ·åºŠãšé床ãå¶åŸ¡ãããã£ãããã¹ã©ã³ãã¹ã±ãŒãã®å±æçãªæ倧å€ããé£ã³åºãããšãã§ãããŸã§ãããã®ãã©ã¡ãŒã¿ãŒãå¢ãããããã£ãããšæ¥æ¿ãªãžã£ã³ããªãã«ã°ããŒãã«ãªæ倧å€ã«ç§»åããããã«ãã©ã¡ãŒã¿ãŒãæžãããŸããã¢ãã«ã«äº€å·®ç¹ãè¿œå ããããšãã§ããŸããããã§ãåå«ã¯äº€é ã«ãã£ãŠåœ¢æããã2ã€ã®ã©ã³ãã ãªèŠªããã©ã³ãã ã«ååã®éºäŒåãåãåããŸãããã®ã¹ããŒã ã§ã¯ãã²ãã ã«ã©ã³ãã ãªçªç¶å€ç°ã®å°å ¥ãæ®ãããšãå¿ èŠã§ãã
ããã§ãã¢ã¬ããµã³ããŒã»ãã«ã³ããšãšã¬ãã»ãã€ãã«ã¯ã®æ¬ããåŒçšããã®ãé©åã ãšæããŸãïŒ
« â , â . , , â . , . , .»
次ã®è³æãç解ããã®ãé£ããå Žåã¯ãåŸã§æ»ã£ãŠãã ãããããã«ãå€ãã®æ©æ¢°åŠç¿ã©ã€ãã©ãªã«ããããã®æ¹æ³ã䜿çšããŠã詳现ã«ç«ã¡å ¥ããã«ãã¬ãŒãã³ã°ãç°¡åã«å®è£ ã§ããŸããé»åã䜿çšããŠã«ãŒããæœåºãããšããéåžžã圌ããããã©ã®ããã«è¡ããã«ã€ããŠã¯ã»ãšãã©é¢å¿ããªããååŸããããã®ãå®å šã«ããç¥ã£ãŠããããã·ã³ã«è«æ±æžãæž¡ããŸãããããããã®æ¹æ³ã«èå³ããã人ã®ããã«ãç°¡åã«æ€èšããŸãã
ãŸããã¢ãã«ã®1ã€ã®åºåãã¥ãŒãã³ã¯1ã€ã®ã¯ã©ã¹ã®ã¿ãæ åœããŸããããããã¥ãŒãã³ãæ åœããã¯ã©ã¹ã®ãªããžã§ã¯ãã§ããå Žåãåºåã§ããã確èªããŸãããã以å€ã®å Žåã¯ãŒãã確èªããŸããå®éã®ã¯ã©ã¹äºæž¬ã§ã¯ããã§ã«ããã£ãŠããããã«ã人工ãã¥ãŒãã³ã¯0ã1ã®ãªãŒãã³ç¯å²ã§ã¢ã¯ãã£ãã«ãªããå€ã¯ããã2ã€ã®æŒžè¿ç·ã«ä»»æã«è¿ã¥ããããšãã§ããŸããããã¯ãã¯ã©ã¹ãããæ£ç¢ºã«æšæž¬ããã»ã©ãå®éã®ã¯ã©ã¹ãšãã®ã¯ã©ã¹ãæ åœãããã¥ãŒãã³ã®æŽ»æ§åãšã®éã®çµ¶å¯Ÿå·®ãå°ãããªãããšãæå³ããŸãã
ããã«ãã£ã®æ°å€ãè¿ãæ倱é¢æ°ãäœæããŠã¿ãŸãããããã¥ãŒã©ã«ãããã¯ãŒã¯ãã¯ã©ââã¹å€ã«è¿ãå€ãçæããå Žåã¯å°ããããã¥ãŒã©ã«ãããã¯ãŒã¯ãã¯ã©ââã¹ã®èª€ã£ãå®çŸ©ã«ã€ãªããå€ãäžããå Žåã¯éåžžã«å€§ãããªããŸãã
å³ 9-ãã¥ãŒãã³ã®åºåã®é¢æ°ãšããŠã®ããã«ãã£é¢æ°ã®ã°ã©ãïŒ1ïŒãªããžã§ã¯ãããã®ã¯ã©ã¹ã«å±ããå ŽåïŒ1ãæåŸ ããïŒã2ïŒãªããžã§ã¯ãããã®ã¯ã©ã¹ã«å±ããªãå ŽåïŒãŒããæåŸ ããïŒ
ããã§ãæ倱é¢æ°ãåŒãšããŠèšè¿°ããŸãããµã€ãºmã®ãã¬ãŒãã³ã°ãµã³ãã«ã®åiçªç®ã®èŠçŽ ã®Yã¯åžžã«0ãŸãã¯1ã®å€ãåãããã2ã€ã®çšèªã®ãã¡1ã€ã ããåžžã«åŒã«æ®ãããšãæãåºããŠãã ããã
æ å ±ã®çè«ã«ç²ŸéããŠãã人ã¯ããã®è¡šçŸã§ã¯ãã¹ãšã³ããããŒãèªèããŸããæ å ±çè«ã®èŠ³ç¹ããèŠããšãåŠç¿ãšã¯ãå®éã®ã¯ã©ã¹ãšã¢ãã«ä»®èª¬ã®éã®çžäºãšã³ããããŒãæå°åããããšã§ãã
æ倱é¢æ°
def binary_crossentropy(X, Y, model): # loss function J = 0 m = len(Y) for i, y in enumerate(Y): J += -(1/m)*(y*np.log(neuron(X[i], model))+(1.-y)*np.log(1.-neuron(X[i], model))) return J
éã¿è¡åã®ä¿æ°ãã©ã³ãã ã«åæåããããšã§ãã¢ãã«ãæ¹åãããã€ãŸãæ倱ãæžããããã«ããããã«å€æŽãå ããããšæããŸããã©ã®ãããã®ééãæ倱é¢æ°ã«åœ±é¿ããããããã£ãŠããå Žåãã©ãã ãã®ééãå€æŽããå¿ èŠãããããããããŸããããã§ãå埮å-åŸé -ã圹ç«ã¡ãŸããé¢æ°ãåŒæ°ã«ã©ã®ããã«äŸåãããã瀺ãã®ã¯åœŒå¥³ã§ããé¢æ°ã1ã€ã®ïŒæ¥µå°ïŒå€ã ãå€åããããã«ãããã€ã®ïŒæ¥µå°ïŒéã«ãã£ãŠåŒæ°ãå€æŽããå¿ èŠããããããããã£ãŠã次ã®ããã«éã¿è¡åãååæåã§ããŸãã
ãã®ã¹ããããç¹°ãè¿ãç¹°ãè¿ããŸããå®éãããã¯ãæ倱é¢æ°ã®å±æçæå°å€ãŸã§ã®ãαã®ãµã€ãºïŒãã®ãã©ã¡ãŒã¿ãŒã¯åŠç¿çãšãåŒã°ããŸãïŒã®å°ããªã¹ãããã§ã®ç·©ãããªåŸé éäžã§ããèšãæãããšãWã®çŸåšã®å€ã«ãã£ãŠå®çŸ©ãããåãã€ã³ãã§ãæ倱é¢æ°ãæãéãå€åããæ¹åãèŠã€ããåŠç¿ã®ãã€ããã¯ã¹ã¯ããŒã«ãå±æçãªæå°å€ã«åŸã ã«è»¢ããã®ã«äŒŒãŠããŸãã
åŸé
éäž
def gradient_descent(model, X_in, Y, number_of_iteratons=500, learning_rate=0.1): X = [[1]+[v.tolist()] for v in X_in] m = len(Y) for it in range(number_of_iteratons): new_model = [] for j, w in enumerate(model): error = 0 for i, x in enumerate(X): error += (1/m) * (neuron(X[i], model) - Y[i]) * X[i][j] w_new = w - learning_rate * error new_model.append(w_new) model = new_model model_loss = binary_crossentropy(X, Y, model) return model
ãšã©ãŒã®éäŒææ³ã¯ãå€å±€ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®å Žåã®ãã®æšè«ã®é£éãç¶ç¶ããŸãã圌ã®ãããã§ãåŸé éäžã«åºã¥ããŠæ·±ãå±€ãèšç·Žããããšãå¯èœã§ããåŠç¿ã¯ãæåŸã®å±€ããæåã®å±€ãŸã§æ®µéçã«è¡ãããŸãããã®æ å ±ã¯ããã®æ¹æ³ã®æ¬è³ªãç解ããã®ã«ååã ãšæããŸãã
ãã¥ãŒã©ã«ãããã¯ãŒã¯ãã¬ãŒãã³ã°ã®äŸ
1ã€ã®ãã©ã¡ãŒã¿ãŒ-幎霢ã«åºã¥ããŠé¡§å®¢ã®è³Œå ¥ã®ç¢ºçãç¥ããããšããŸããè³Œå ¥ã®ç¢ºçã50ïŒ ãè¶ ããå Žåã«è奮ãããã¥ãŒãã³ãäœæããŸãã
å³ 10-è³Œå ¥ãã人ãšè³Œå ¥ããªãã£ã人ã®ãã¬ãŒãã³ã°ãµã³ãã«
ãããã£ãŠããã¥ãŒãã³ã«ã¯ã¯ã©ã€ã¢ã³ãã®å¹Žéœ¢ã«é¢é£ãã1ã€ã®å容äœããããŸããããã«ãã·ããïŒãŸãã¯ãªãã»ããïŒãæ åœãã1ã€ã®ãã€ã¢ã¹ã¡ã³ããŒãè¿œå ããŸããããšãã°ããããã®ã»ããã¯ç·åœ¢ã«åé¢ã§ããŸããããããããã®å¢çãã€ãŸããæé«ã®åé¢è¶ æ²é¢ïŒ1次å ã®å Žå-ãã€ã³ãïŒã¯ã42æ³ã§ãããã®éã«äœçœ®ããŸãã
ãã¥ãŒã©ã«ãããã¯ãŒã¯ã§ã¯ã42æ³ã§0.5æªæºãé«éœ¢ã®é¡§å®¢ã§ã¯0.5ãè¶ ããè³Œå ¥ã®ç¢ºçãäžããå¿ èŠããããŸããã¢ã¯ãã£ããŒã·ã§ã³é¢æ°ãåŒã³åºããšãæ£ã®åŒæ°ã«å¯ŸããŠ0.5ãã倧ããå€ãè¿ããè² ã®åŒæ°ã«å¯ŸããŠ0.5ããå°ããå€ãè¿ããŸãããã®ããããã®ã¢ã¯ãã£ããŒã·ã§ã³é¢æ°ããããããå€ã«ã·ããããæ©èœãå¿ èŠã§ããåæã«ãç¹åŸŽãã¯ãã«xã®é¢æ°ãšããŠã®åãã¥ãŒãã³ã®å±èµ·ã®çšåºŠã¯éã¿è¡åã®ä¿æ°ã«äŸåãããããã£ãŠããã®ãããªç¹åŸŽãã¯ãã«ãæã€èŠçŽ ããã®ã¯ã©ã¹ã«å±ãã確çã«äŸåããããããã¬ãŒãã³ã°ãµã³ãã«ã«æé©ãªæŽ»æ§åé¢æ°ã®ãã®ãããªç Žå£çãäºæ³ãããŸãã
å³ 11-ã¯ã©ã€ã¢ã³ãã®å¹Žéœ¢ã«å¯Ÿãããã¥ãŒãã³ã®äºæ³ãããå¿çãçµæã®ã確信床ãã¯ããŸããŸã§ãåŒæ°ã®ä¿æ°ã«ãã£ãŠèª¿æŽãããŸãã
ãããæ°åŠçã«èšè¿°ããéã¿ãããªãã¯ã¹ã«ãã1ã€ã®ãã€ã¢ã¹é ãå¿ èŠãªçç±ãç解ããŸããé¢æ°fïŒxïŒãäŸãã°42ã ãå³ã«ã·ããããã«ã¯ãåŒæ°fïŒx-42ïŒãã42ãåŒãå¿ èŠããããŸããåæã«ãäŸãã°0.25ãåŒæ°ã«æããŠæ¬¡ã®é¢æ°fïŒ0.25ïŒx-24ïŒïŒãååŸããããšã«ãããé¢æ°ã®åŒ±ãå€æ²ãååŸããããšæããŸããæ¬åŒ§ãéããšã次ã®ããã«ãªããŸãã
ãã®å Žåãéã¿è¡åã®æãŸããä¿æ°ã¯w = 0.25ã§ãã·ããã¯b = -10.5ã§ãããã ããbãéã¿è¡åã®ãŒãä¿æ°ïŒw 0 = bïŒã§ãããšä»®å®ã§ããŸããããšãã°ããŒãèšå·ãåžžã«åäžïŒx 0= 1ïŒã次ã«ãããšãã°ãx ïŒ15ïŒ = {x ïŒ15ïŒ 0ãx ïŒ15ïŒ 1 } = [1ã30]ã§è¡šããã45æ³ã®15çªç®ã®ããã¯ãã«åãããã顧客ã¯ã68ïŒ ã®ç¢ºçã§è³Œå ¥ã§ããŸãããããã®ãã¹ãŠã®ä¿æ°ã¯ããã®ãããªåçŽãªäŸã§ãããç®ã§ãææ¡ããã®ã¯å°é£ã§ãããããã£ãŠãå®éã«ã¯ãæ©æ¢°åŠç¿ã¢ã«ãŽãªãºã ã«å¯Ÿãããããã®ãã©ã¡ãŒã¿ãŒã®æ€çŽ¢ãä¿¡é ŒããŠããŸãããã®äŸã§ã¯ãéã¿è¡åã®2ã€ã®ä¿æ°ïŒw 0 = bããã³w 1ïŒãæ¢ããŠããŸãã
ã¢ãã¡ãŒã·ã§ã³1-æ£èŠåãªãã®ããŒã¿ã®é²åçåŠç¿
éã¿è¡åã®ä¿æ°ãã©ã³ãã ã«åæåããé²åã¢ã«ãŽãªãºã ã䜿çšããããšã«ããã100äžä»£åŸã«èšç·Žããããã¥ãŒã©ã«ãããã¯ãŒã¯ãåŸãŸãããåžæããåŠç¿ææãããéãããã®ãããªæ¥æ¿ãªãžã£ã³ããªãã§ååŸããã«ã¯ããã¬ãŒãã³ã°ã®åã«ããŒã¿ãæ£èŠåããå¿ èŠããããŸãã
ã¢ãã¡ãŒã·ã§ã³2-æ£èŠåããŒã¿ã®é²ååŠç¿
åŸé éäžæ³ã¯æãæ£ç¢ºã«æ©èœããŸãããã®æ¹æ³ã䜿çšããå ŽåãããŒã¿ã¯åžžã«æ£èŠåããå¿ èŠããããŸããé²åçã¢ã«ãŽãªãºã ãšã¯å¯Ÿç §çã«ãåŸé éäžæ³ã¯ãçªç¶å€ç°ãã«é¢é£ãããžã£ã³ããçµéšããŸããããåŸã ã«æé©ã«ç§»åããŸãããããããã€ãã¹ã¯ããã®ã¢ã«ãŽãªãºã ã極å°å€ã«ãšã©ãŸããããããæãåºããªãããŸãã¯åŸé ãå®è³ªçã«ãæ¶å€±ãããŠåŠç¿ãåæ¢ããå¯èœæ§ãããããšã§ãã
ã¢ãã¡ãŒã·ã§ã³3-åŸé éäžæ³ã«åºã¥ãåŠç¿
è³Œå ¥ããã¯ã©ã¹ãšè³Œå ¥ããŠããªãã¯ã©ã¹ã®ã»ãããç·åœ¢åé¢å¯èœã§ããå Žåããã¥ãŒãã³ã¯ãã®æ±ºå®ã«ããèªä¿¡ãæã¡ããã®æŽ»æ§åã®çšåºŠãå€æŽãããšããããã®ã»ããã®å¢çã§ããé¡èãªéªšæãçããŸãã
ã¢ãã¡ãŒã·ã§ã³4-ç·åœ¢åé¢å¯èœã»ããã®åŠç¿
äžèšã«åºã¥ããŠã以äžãå«ãèšç®ã°ã©ãã®åœ¢åŒã®å€å žçãªãã¥ãŒã©ã«ãããã¯ãŒã¯ãæ³åã§ããŸãã
- å ¥åé ç¹x;
- åºåaã®å€ãæã€ãã¥ãŒãã³ã§ããé ç¹ã
- ãã€ã¢ã¹bã®åå ãšãªãé ç¹ã
- åã®å±€ã®åºåå€ã«ãéã¿è¡åwã®å¯Ÿå¿ããä¿æ°ãæãããšããžã
- 仮説h wãbïŒxïŒã¯ãæåŸã®å±€ã®åºåã®çµæã§ãã
å³ 12-å€å žçãªäººå·¥ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®ã°ã©ãã®èšç® TensorFlow
ã©ã€ãã©ãªã®ãªã³ã©ã€ã³ãµã³ãããã¯ã¹ã®äŸãããã€ãèããŠã¿ãŸãããããã¹ãŠã®äŸã§ããªããžã§ã¯ããå¹³é¢äžã«ãã2ã€ã®ã¯ã©ã¹ãåé¢ããå¿ èŠããããŸããå ¥åã¬ã€ã€ãŒã«ã¯ã暪座æšãšçžŠåº§æšã«æ²¿ã£ããªããžã§ã¯ãã®åº§æšã«å¯Ÿå¿ããå€ãæã€2ã€ã®ãå容äœãããããŸãïŒããã«1ã€ã®ãã€ã¢ã¹ããã€ã¢ã¹ã¯ã¢ãã¡ãŒã·ã§ã³ã«è¡šç€ºãããŸããïŒãåè¿°ã®ããã«ãç·åœ¢ã«åé¢å¯èœãªã»ããã§ãã¬ãŒãã³ã°ãè¡ãã«ã¯ãåºåãã¥ãŒãã³ã1ã€ã ãããã°ååã§ããé ããïŒé£æ³çãªïŒå±€ã¯ãããŸãããåŠç¿ã¯ãéäŒææ³ã«åºã¥ããŠããŸããã¢ãã¡ãŒã·ã§ã³5-1ã€ã®åºåãã¥ãŒãã³ãåå²ç·ãèŠã€ãã
ã¿ã¹ã¯ãè€éã«ããã»ãããåé¢ããŠã¿ãŸããããæåã®èŠçŽ ã¯äž¡æ¹ã®åº§æšã§æ£ãŸãã¯è² ã®å€ã®ã¿ãæã¡ã2çªç®ã®èŠçŽ ã¯åº§æšã®1ã€ã®æ£ãš1ã€ã®è² ã®å€ãæã¡ãŸãããã®äŸã§ã¯ã1ã€ã®åå²ç·ãçãããšãã§ããªããããé ãã¬ã€ã€ãŒãå¿ èŠã§ããæå°éããå§ããŠãé ãå±€ã«2ã€ã®ãã¥ãŒãã³ãè¿œå ããŠã¿ãŸãããã
ã¢ãã¡ãŒã·ã§ã³6-2ã€ã®é£æ³ãã¥ãŒãã³ãš2ã€ã®åå²ç·
ã芧ã®ãšãããé ãå±€ã®2ã€ã®ãã¥ãŒãã³ã¯ãã®ã¿ã¹ã¯ã«å¯Ÿå¿ããŠããŸãããæè¯ã®æ¹æ³ã§ã¯ãããŸãããåŠç¿ããã»ã¹ã§ãã¥ãŒãã³ã®ç¹æ®åïŒååïŒãã©ã®ããã«çºçãããã«æ³šæããŠãã ããã次ã«ã4ã€ã®ãã¥ãŒãã³ã§æ§æãããé ãã¬ã€ã€ãŒãäœæããŸãã
ã¢ãã¡ãŒã·ã§ã³7-4ã€ã®é£æ³ãã¥ãŒãã³ãš4ã€ã®åå²ç·
ãã¥ãŒã©ã«ãããã¯ãŒã¯ã¯ãããããŸãåŠçããŸããããã¬ãŒãã³ã°ã®å®æœæ¹æ³ã«æ³šæããŠãã ãããæåã«ããã¥ãŒã©ã«ãããã¯ãŒã¯ã¯æãåçŽãªãœãªã¥ãŒã·ã§ã³-åå²å»äžãèŠã€ããŸããããã®åŸããã¥ãŒãã³ã®åå°éåããããŸãããçŸåšã§ã¯ãåé衚瀺ïŒé£æ³ïŒãã¥ãŒãã³ããã®çãã»ã°ã¡ã³ããæ åœããŠããŸãã
ããªãè€éãªåé¡-ç°ãªãã¹ãã€ã©ã«ã¢ãŒã ã«ãã2ã€ã®ã»ããã®èŠçŽ ã®åé¢-ã解決ããŠã¿ãŸãããã
ã¢ãã¡ãŒã·ã§ã³8-ããã«ããã¯ããããžãåããå€å±€ãã¥ãŒã©ã«ãããã¯ãŒã¯
è€éãªåé¡ã解決ããã«ã¯ãå€ãã®é ãå±€ãå¿ èŠã§ãããã¥ãŒã©ã«ãããã¯ãŒã¯ã¯ãæåã®é ãå±€ããæåŸã®å±€ãŸã§ãã¥ãŒãã³ã®æ°ãæžå°ããããã«ããã¯ããããžã«ããŸã察å¿ããŸããé£æ³ãã¥ãŒãã³ãç¹æ®åãããšãã«çããè€éãªãã¿ãŒã³ã«æ³šæããŠãã ããããã£ãŒããã¥ãŒã©ã«ãããã¯ãŒã¯ã®å Žåãé ããã¥ãŒãã³ã«ã¯ReLUïŒä¿®æ£ç·åœ¢ãŠãããïŒã¢ã¯ãã£ããŒã·ã§ã³é¢æ°ã䜿çšããæåŸã®å±€ã«ã¯éåžžã®ããžã¹ãã£ãã¯ã¢ã¯ãã£ããŒã·ã§ã³ïŒçæ³çã«ã¯ãœããããã¯ã¹ã¢ã¯ãã£ããŒã·ã§ã³ïŒã䜿çšããããšããå§ãããŸãã
ããã§ã人工ç¥çµå路網ã®è¶ çæã³ãŒã¹ãçµäºããç¥èãå®è·µã«ç§»ãããšãã§ãããšæããŸããããããããã°ã©ãã³ã°èšèªã§å©çšã§ããããã«ãªã£ãæ¢è£œã®ã©ã€ãã©ãªã«ã¢ãã«ãæ§ç¯ãããã®æ¹åã§çè«çãªç¥èãåŸã ã«æ·±ããŠããããšããå§ãããŸãã
ã¢ãã«ãã¬ãŒãã³ã°
ãã¬ãŒãã³ã°ãµã³ãã«ãšçè«çç¥èã®äž¡æ¹ãåŸãããããã¢ãã«ã®ãã¬ãŒãã³ã°ãéå§ã§ããŸãããã ããåé¡ã¯ãã»ããã®èŠçŽ ãäžåçãªæ¯çã§è¡šç€ºãããããšãå€ãããšã§ããè³Œå ¥è ã¯5ïŒ ãéè³Œå ¥è ã¯95ïŒ ã§ããããã§ã¯ã©ããã£ãŠèšç·ŽããŸããïŒçµå±ã®ãšããã誰ãè³Œå ¥ããªããšäž»åŒµããããšã§ã95ïŒ ã®ä¿¡é Œæ§ãéæã§ããŸãã
ããããããã®å Žåã®ç²ŸåºŠã¡ããªãã¯ã¯ç°ãªãããã¬ãŒãã³ã°ãåççã§ããå¿ èŠããããŸããããã«ããããã¥ãŒã©ã«ãããã¯ãŒã¯ãåãæããã«ééã£ãçµè«ãåºããªãããã«ããŸãããã®ããã«ãç°ãªãã¯ã©ã¹ã®åæ°ã®èŠçŽ ãå«ããã¬ãŒãã³ã°äŸã§ãã¥ãŒã©ã«ãããã¯ãŒã¯ãããã£ãŒããããããšãææ¡ããŸãã
ããšãã°ããµã³ãã«ã20,000åãããªãããã®ãã¡1,000åãè³Œå ¥ããŠããå Žåãåã°ã«ãŒããã500åã®ãµã³ãã«ãã©ã³ãã ã«éžæãããã¬ãŒãã³ã°ã«äœ¿çšã§ããŸãããããŠããã®æäœãäœåºŠãç¹°ãè¿ããŸããããã¯åŠç¿ããã»ã¹ã®å®è£ ãå°ãè€éã«ããŸãããæèœãªã¢ãã«ãåŸãã®ã«åœ¹ç«ã¡ãŸãã
ã¢ãã«ãšãã¬ãŒãã³ã°ã¢ã«ãŽãªãºã ãéžæãããããµã³ãã«ãéšåã«åå²ããããšããå§ãããŸããå šäœã®70ïŒ ãå ãããã¬ãŒãã³ã°ãµã³ãã«ã§ãã¬ãŒãã³ã°ãè¡ããçµæã®ã¢ãã«ã®å質ãåæããããã«å¿ èŠãªãã¹ããµã³ãã«ã«30ïŒ ãå¯ä»ããŸãã
ã¢ãã«å質è©äŸ¡
ã¢ãã«ãæºåãããããã®å質ãé©åã«è©äŸ¡ããå¿ èŠããããŸãã ãããè¡ãããã«ã次ã®æŠå¿µã玹ä»ããŸãã
- TPïŒçéœæ§ïŒ-çéœæ§ãåé¡åã¯ã顧客ãè³Œå ¥ããããšã決å®ãã圌ã¯è³Œå ¥ããŸããã
- FPïŒFalse PositiveïŒ-誀æ€ç¥ãåé¡åã¯é¡§å®¢ãè³Œå ¥ããããšã決å®ããŸãããã圌ã¯è³Œå ¥ããŸããã§ãããããã¯ããããã第1çš®ã®ãšã©ãŒã§ããç¹ã«åé¡åãäœããã®ç æ°ã®æ€æ»ã§ããå Žåã第2çš®ã®ééãã»ã©æãã¯ãããŸããã
- FNïŒFalse NegativeïŒ-åœé°æ§ãåé¡åã¯ãã¯ã©ã€ã¢ã³ãã¯è³Œå ¥ããªããšæ±ºããŸããããè³Œå ¥ããïŒãŸãã¯æ¢ã«è³Œå ¥ããïŒããšãã§ããŸããããã¯ããããã第2çš®ã®ééãã§ããéåžžãã¢ãã«ãäœæãããšãã¯ã2çªç®ã®ãšã©ãŒãæå°éã«æããããã«ãã£ãŠç¬¬1çš®ã®ãšã©ãŒãå¢å ãããããšãæãŸããã§ãã
- TNïŒTrue NegativeïŒ-æ¬åœã«ãã¬ãã£ããåé¡åã¯ãã¯ã©ã€ã¢ã³ããè³Œå ¥ããªããšå€æããè³Œå ¥ããŸããã§ããã
ããŒã»ã³ãä¿¡é Œæ§ã®çŽæ¥è©äŸ¡ã«å ããŠãäžèšã®ãã€ããªåé¡çµæã«åºã¥ã粟床ïŒè±èªã®ç²ŸåºŠïŒãå®å šæ§ïŒè±èªã®åçŸçïŒãªã©ã®ã¡ããªãã¯ããããŸãã
å³ 13-ç°ãªãã¢ãã«ã®åé¡çµæã®æ¯èŒ
ã芧ã®ãšããããã¥ãŒã©ã«ãããã¯ãŒã¯ã¯ã以åã®ã¹ã³ã¢ãªã³ã°ã¢ãã«ãšæ¯èŒããŠã第2çš®ã®ãšã©ãŒã3åå°ãªããéåžžã«åªããŠããŸãã確ãã«ãæœåšçãªé¡§å®¢ãéããšãæœåšçãªå©çãéããŸããããã§ãå©çšå¯èœãªããŒã¿ã«åºã¥ããŠã¢ãã«ã®å質ææšãå°ãåºããŸãããã
ä¿¡é Œæ§ææš
æãåçŽãªã¡ããªãã¯ã¯ã粟床ã¡ããªãã¯ã§ãããããããã§ã«ããã£ãŠããããã«ããã®ã¡ããªãã¯ã¯ã¢ãã«ã®å¯äžã®ã¡ããªãã¯ã§ã¯ãããŸãããç¹ã«ããµã³ãã«ã«ãã€ã¢ã¹ãããå Žåãã€ãŸããç°ãªãã¯ã©ã¹ã®ä»£è¡šãç°ãªã確çã§åºäŒãå Žåã
æ£ç¢ºããšå®å šæ§
粟床ã¯ãã¯ã©ã¹ã®ãªããžã§ã¯ããšããŠå®çŸ©ãããã¹ãŠã®ãªããžã§ã¯ãã«å¯Ÿãããã¯ã©ã¹ã®æ£ããæšæž¬ããããªããžã§ã¯ãã®æ¯çã瀺ããŸããããšãã°ã115ãè³Œå ¥ãããã®ãã¡37ãå®éã«è³Œå ¥ããããšã«ããŸãããã€ãŸãã粟床ã¯0.33ã§ããå®å šæ§ïŒè±èªã®ãªã³ãŒã«ïŒã¯ãã¯ã©ã¹ã®ãã¹ãŠã®ä»£è¡šã«å¯Ÿããã¯ã©ã¹ã®æ£ããæšæž¬ããããªããžã§ã¯ãã®æ¯çã瀺ããŸããããšãã°ãç§ãã¡ã®äžã§å®éã«è³Œå ¥ããã®ã¯37åã§ãè³Œå ¥ããã®ã¯43åã ãã ã£ãããããã«ãã¹ã¯0.88ã§ãã
å³ 14-ãšã©ãŒè¡šãŸãã¯æ··åãããªãã¯ã¹
Fã¡ãžã£ãŒ
Fã¡ãžã£ãŒïŒEngãF1ã¹ã³ã¢ïŒããããŸã-é«èª¿æ³¢ã®ç²ŸåºŠãšå®å šæ§ã®å¹³åã§ããåäžã®æ°å€æž¬å®ã䜿çšããŠã¢ãã«ãæ¯èŒããã®ã«åœ¹ç«ã¡ãŸãã
ããããã¹ãŠã®ã¡ããªãã¯ã䜿çšããŠãã¢ãã«ãè©äŸ¡ããŸãã
å³ 15-ããŸããŸãªçµ±èšææšã«åºã¥ããã¢ãã«ã®å質ã®è©äŸ¡
å³ãããããããã«ãã¢ãã«ã®å質ã®æ倧ã®åãã¯å®å šæ§ã®ææšã«æ£ç¢ºã«ãããŸãïŒè±èªã®ãªã³ãŒã«ïŒããã¥ãŒã©ã«ãããã¯ãŒã¯ã¯ãæœåšçãªé¡§å®¢ã®88ïŒ ãæšæž¬ããããã12ïŒ ãéããŠããŸããå€ãã¹ã³ã¢ãªã³ã°ã¢ãã«ã§ã¯ãæœåšçãªé¡§å®¢ã®36ïŒ ãéãããããŒãžã£ãŒã«ã¯64ïŒ ããæž¡ãããŸããã§ããããããå®éã«ãã¹ã³ã¢ãªã³ã°ã«åœ±é¿ãäžããããŸããŸãªåçã®éèŠåºŠã®ä¿æ°ãéžæããããã«ãã¥ãŒã©ã«ãããã¯ãŒã¯ãå§ä»»ããã»ããããçç±ã§ããçµå±ã®ãšããããã·ã³ã¯ã¡ã¢ãªå šäœã«ãµã³ãã«å šäœãä¿æãããã®äžã®ãã¿ãŒã³ãèŠã€ããåªããäºæž¬èœåãæã€ã¢ãã«ãæ§ç¯ããããšãã§ããŸãã
ã¢ãã«ã®è§£é
æ¢è£œã®ã¢ãã«ãããã°ãå質åæããåŸããã粟床ãæåŸ ããŠäœ¿çšã§ããŸããè€éãªïŒå€å±€ïŒã¢ãã«ãããã»ã¹ã«å°å ¥ã§ããå Žåãããã¯é©åã§ãããããã§ãªãå Žåã¯ãåäžå±€ãã¥ãŒã©ã«ãããã¯ãŒã¯ããéåžžã®ã¹ã³ã¢ãªã³ã°ã¢ãã«ãååŸã§ããŸãããã®ãããç§ãã¡ã¯åœŒå¥³ã®ããŒãã倧èã«èª¿ã¹ãããã«ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®æ§ç¯ã«ç²ŸéããŠããŸãã
ç·åœ¢ã¢ãã«ã®ã¹ã³ã¢ãªã³ã°åŒãš1ã€ã®ãã¥ãŒãã³ïŒãŸãã¯åå±€ãã¥ãŒã©ã«ãããã¯ãŒã¯ïŒã®é¢æ°ãæ¯èŒããŸãã
ã¢ã¯ãã£ããŒã·ã§ã³é¢æ°ã®åŒæ°ã§ããåŒã¯ãç·åœ¢ã¹ã³ã¢åŒãšåäžã§ããããšãããããŸãããããã£ãŠãåå±€ãã¥ãŒã©ã«ãããã¯ãŒã¯ããéã¿è¡åã®å€ãããã«ãããããšã«ãããããããã¹ã³ã¢ãªã³ã°ã¢ãã«ä¿æ°ãšããŠäœ¿çšã§ããŸããçŸåšã®ã¿ããããã®ä¿æ°ã¯ã倧éã®ããŒã¿ã«åºã¥ããŠã¢ã«ãŽãªãºã ã«ãã£ãŠæ éã«éžæãããŸãã
次ã«ããã¥ãŒã©ã«ãããã¯ãŒã¯ã®å°å ¥ååŸã®ä¿æ°ã«åºã¥ããŠç·åœ¢ã¹ã³ã¢ãªã³ã°ã®çµæãæ¯èŒããŸããããžã¹ãã£ãã¯æŽ»æ§åé¢æ°ãå€ãäžããããšãæãåºããŠãã ããïŒãã¥ãŒã©ã«ãããã¯ãŒã¯ã®èŠ³ç¹ãã-è³Œå ¥ãã人ã ã®ã¯ã©ã¹ã«å±ãã確çïŒã¯ãåŒæ°ã®æ£ã®å€ïŒéã¿è¡åã«åºã¥ãã¹ã³ã¢ïŒã§0.5ãè¶ ããŸãããã€ã³ããã¹ã±ãŒãªã³ã°ããããã«ããã¥ãŒã©ã«ã¹ã³ã¢ãªã³ã°ã®å€ã«100ãæããŠããããå€ãšããŠ500ãè¿œå ããŸãããå€ãã¹ã³ã¢ãªã³ã°ã§ã¯ããããå€ã¯æåã§éžæãããåèšã§170ã«ãªããŸããããããã¯ãã¹ãŠãã¢ãã«èªäœã«åœ±é¿ãäžããªãåãªãç·åœ¢æäœã§ãã
å³ 16-å€ãã¹ã³ã¢ãªã³ã°ã¢ãã«å ã§è³Œå ¥ããïŒèµ€ïŒããã³è³Œå ¥ããªãã£ãïŒéïŒ
顧客ã®ååžãååžãããããããã«ã顧客ã¯ã¹ã³ã¢å€ã®å šç¯å²ã«åºãããããŠããŸããå®å šæ§ïŒè³Œå ¥ãã顧客ã®ç·æ°ã®ã¢ãã«ã«ãã£ãŠäºæž¬ãããå²åïŒã¯64ïŒ ã§ãã
å³ 17-ãã¥ãŒã©ã«ã¹ã³ã¢ãªã³ã°ã¢ãã«ã®ãã¬ãŒã ã¯ãŒã¯å ã§è³Œå ¥ããïŒèµ€ïŒããã³è³Œå ¥ããªãã£ãïŒéïŒé¡§å®¢ã®ååžã
ååžããããã¥ãŒã©ã«ãããã¯ãŒã¯ãå€ãã¢ãã«ãããè³Œå ¥ãããŠãŒã¶ãŒãšè³Œå ¥ããªãã£ããŠãŒã¶ãŒãåé¢ã§ããããšãããããŸããã»ãšãã©ã®å Žåãè³Œå ¥ããªãã£ã人ã¯ãããå€æªæºã®å€ãåãåããè³Œå ¥ãã人ã¯äžèšã®å€ãåãåããŸãããå®å šæ§ïŒè³Œå ¥ãã顧客ã®ç·æ°ã®ã¢ãã«ã«ãã£ãŠäºæž¬ãããã·ã§ã¢ïŒã¯88ïŒ ã§ãã
çµæ
åé¡ã解決ããããã«ãé¢çšã倧éã«è³Œå ¥ãã人ãã¡ã«ã§ããã ãå€ãã®æéãè²»ããããã£ãã®ã§ããããã«ãæãå®ãã¬ãŒããè³Œå ¥ãã顧客ãåæ Œç¹ãç²åŸããªãã¹ã³ã¢ãªã³ã°ã¢ãã«ãäœæããããšèããŸããã
ä»åã¯ããã¥ãŒã©ã«ãããã¯ãŒã¯ããã¬ãŒãã³ã°ããããã«ããµã³ãã«ãäžå®ä»¥äžã®éé¡ã§è³Œå ¥ãã人ã®ã¯ã©ã¹ãšãäœé¡ã§è³Œå ¥ããããŸã£ããè³Œå ¥ããªãã£ãã¯ã©ã¹ã«åããŸããã
å³ 18-æçµçãªãã¥ãŒã©ã«ã¹ã³ã¢ãªã³ã°ã¢ãã«å ã§ã®è³Œå ¥æžã¿ïŒèµ€ïŒããã³æªè³Œå ¥ïŒéïŒã®é¡§å®¢ã®ååžã
ä»ã®ã»ãŒåçã®ç²ŸåºŠã€ã³ãžã±ãŒã¿ãŒã䜿çšããŠããã¥ãŒã©ã«ãããã¯ãŒã¯ã¯ããé«ãå®å šæ§ãéæã§ããé«äŸ¡ãªæéãã©ã³ãè³Œå ¥ãã人ã®87ïŒ ãã«ããŒã§ããŸãããæ¯èŒã®ããã«ïŒå€ãã¹ã³ã¢ãªã³ã°ã¯77ïŒ ãã管çããŠããŸãããããã¯ãå°æ¥çã«éèŠãªæœåšé¡§å®¢ã®10ïŒ ãã«ããŒã§ããããšãæå³ããŸããåæã«ãåŸç¹ã®é«ãé¢çšããé«äŸ¡ãªé¢çšãè³Œå ¥ãã人ã®å²åã¯ã»ãŒåãã§ãããã¥ãŒã©ã«ãããã¯ãŒã¯ãšå€ãã¢ãã«ã§ã¯ãããã23ïŒ ãš24ïŒ ã§ããã¹ã³ã¢ãªã³ã°å€ã¯è³Œå ¥éé¡ãšããçžé¢ããŠããããšãããããŸãã
å³ 19-å€ãã¹ã³ã¢ãªã³ã°ã¢ãã«ãšæ°ããã¹ã³ã¢ãªã³ã°ã¢ãã«ã®å質ã®æ¯èŒ
ãã®èšäºã§ã¯ã次ã®ããšãè¡ããŸãã
- ããŒã¿ãã€ãã³ã°ã®ãã¹ãŠã®äž»èŠãªæ®µéã調ã¹ãŸããã
- ããŒã¿ã®æºåãšãã¬ãŒãã³ã°ã®äž¡æ¹ã§ãå€ãã®æçšãªãã¯ããã¯ãåŠã³ãŸããã
- å€å žçãªäººå·¥ãã¥ãŒã©ã«ãããã¯ãŒã¯ã®çè«ã«ç²ŸéããŠããã
- ã¢ãã«å質ã®åæã«å¯ŸããããŸããŸãªçµ±èšçã¢ãããŒããæ€èšããŸããã
- ç·åœ¢ã¹ã³ã¢ãªã³ã°ã¢ãã«ã®æ§ç¯äŸã䜿çšããŠããã¥ãŒã©ã«ãããã¯ãŒã¯ã®äœæããåæãå®è£ ãŸã§ã®ãã¹ãŠã®æ®µéã説æããŸããã
- ææ°ã®æ©æ¢°åŠç¿ã¢ã«ãŽãªãºã ãå®éã®ããžãã¹äžã®åé¡ã®è§£æ±ºã«ã©ã®ããã«åœ¹ç«ã€ãã瀺ããŸããã
質åãææ¡ãã³ã¡ã³ããããå Žåã¯ãã³ã¡ã³ãã§è°è«ããŸãããã