Habréã®ããã°ã§ã¯ãéèåžå Žã®ããŸããŸãªæè¡ãæ€èšããã ãã§ãªããåæã®éçšã§ã¢ããªã¹ãã䜿çšããããŸããŸãªããŒã«ã«ã€ããŠã説æããŠããŸãã ç¹ã«ãããã»ã©åã§ã¯ãªãããã©ã³ãã ãŠã©ãŒã¯ä»®èª¬ãéèåžå Žã®ç¶æ ãäºæž¬ããããã«ã©ã®ããã«äœ¿çšã§ãããã«ã€ããŠæžããã NMRQLãããžãã¡ã³ãã®å®éã¢ããªã¹ãã§ããStuart Reedã¯ããã¥ãŒãªã³ã°ãã¡ã€ãã³ã¹ã®Webãµã€ãã§åœŒã®ç 究çµæãå ¬éããåžå Žä»®èª¬ã®ã©ã³ãã æ§ããã¹ãããããã«ãã®ä»®èª¬ãé©çšããŸããã
ããã¯ãä¹±æ°çæåšãNISTãã¹ãã°ã«ãŒãããå®è¡ãããŠãå©çã®ããã«åžå Žã®éå¹çæ§ãæªçšããè匱æ§ãã©ãã§çºçããããç解ãããšãããã®ã§ããã å®éšäžã«ãèè ã¯ãäžéšã®è©å€ã®é«ãç§åŠè ãä¿¡ããŠããããã«ãåçŽãªã³ã€ã³ãã¹ã®èŠ³ç¹ããåžå Žè¡åã説æã§ããªããšããçµè«ã«éããŸããã äžéšã®ãã¹ãã§ã¯ãåžå Žã®æ¯ãèãã«ãããäžå®ã¬ãã«ã®ããã€ãºããä¿®æ£ããããšãã§ããŸããã ãããã®1ã€-ç·åœ¢è€é床ãã¹ã-ã¯ãã©ã³ãã æ§ãšå§çž®åºŠã®é¢ä¿ã®èããæãåºããããããèè ã®æ³šæãåŒããŸããã
æ°ããèšäºã§ãReid ã¯ãããŒã¿å§çž®ã¢ã«ãŽãªãºã ã以åã«èª²ãããã¿ã¹ã¯ã«ããããå©ç¹ãç解ããããšããŸããã ç§ãã¡ã¯ããªãã®æ³šæã«ãã®äœåã®é©å¿ããã翻蚳ãæ瀺ããŸãã
åžå Žããã©ãŒãã³ã¹ãšã©ã³ãã ãŠã©ãŒã¯ä»®èª¬
ãã®æ¥µç«¯ãªè¡šçŸã§ã¯ãåžå Žå¹ç仮説ã¯ãäŸ¡æ Œã®äŸ¡å€ã«ãããŠããã¹ãŠã®æ å ±ããå ¬çã§ããããšç§çå人ã«ãšã£ãŠå ¥æå¯èœã§ããããšãå³åº§ãã€æ£ç¢ºã«èæ ®ããããšè¿°ã¹ãŠããŸãã ãããã£ãŠãçŸåšã®çžå Žã¯çŸåšã®ç¶æ³ãæ£ç¢ºãã€ããªãå®å šã«åæ ããŠãããããææ¥ã®äŸ¡æ Œäºæž¬ã¯åžå Žã®çŸåšã®ç¶æ ã«åºã¥ããŠããå¿ èŠããããŸãã
ãã®çè«ã«ãããšãéå»ã®è©³çŽ°ãªçµæžåæããã¡ã³ãã¡ã³ã¿ã«åæã¯ãå°æ¥ã®ããæ£ç¢ºãªäºæž¬ãæäŸããŸããã æ å ±ã®å®å šæ§ãšååæ§ãšããç¹ã§å¹ççãªåžå Žã§ã¯ããææ¥ãã«èŠããæªç¥ã®ããŒã¿ã«äŸåãããããæªæ¥ã¯å®çŸ©ãããŠããŸããã ä»æ¥ã®ãªãã¶ãŒããŒã®äœçœ®ã«åºã¥ããŠãå°æ¥ã®å€æŽã¯ã©ã³ãã ã§ãã æéå ã«æ è¡ããæ¹æ³ãç¥ã£ãŠããã¢ããªã¹ãã®ã¿ãåžå Žãããããã³ã°ãã§ããŸãã
ãã®ä»®èª¬ã«ã¯ãç¶æ³ã®ããè€éãªç解ãé²åçããã³è¡åçèŠå ã®åœ±é¿ãæå±ãããšããŠãããã®ã¡ãªããããããŸãã ãã®çè«ã®ã¡ãªããã¯åŠå®ã§ããŸããã ãŸããåžå Žã®åããäºæž¬ããã®ã¯æ¬åœã«é£ããã§ãã 第äºã«ãå¹³åçãªãã¬ãŒã€ãŒãåžå Žãç Žãããšãã§ããäžé£ã®ã¢ã¯ã·ã§ã³ãæ¬åœã«ãããŸããã 第äžã«ãå¹çæ§ä»®èª¬ã¯ãäŸãã°ããªããã£ããªã©ã®éèé©æ°ã®åºçŸãéåžžã«è«ççã«èª¬æããŸãã
å¹çæ§ä»®èª¬ã«å¯Ÿããæãæ·±å»ãªæ¹å€ã¯ãè¡åã®è²¡ååæããæ¥ãŠãããããã¯ãã®åºæ¬çãªä»®èª¬ã«çåãæããããŠããŸãã ãã®ã¢ãããŒãã®äžç°ãšããŠãæè³å®¶ã¯åççã«è¡åãããæ ªåŒã«ãå ¬æ£ãªãäŸ¡æ Œãäžããããšãã§ããªããšäž»åŒµãããŠããŸãã å®éã«ã¯ããã¬ã€ã€ãŒã¯äŸ¡æ Œã«ã€ããŠäžè¬çãªæåŸ ãæã¡ãŸããã
åžå Žéçºã®é²åè«ã¯ãããã«ããã€ãã®è°è«ãè¿œå ããŸãã 圌女ã¯ããã®ã·ã¹ãã ã¯ãé²åã®æ³åã«åŸã£ãŠãçžäºäœçšããç°çš®ã®ãšãŒãžã§ã³ãã§æ§æãããŠãããšèããŠããŸãã ãã®çžäºäœçšã§ã¯ãã©ã³ãã æ§ãå¢ãã䟡å€ãè³ç£å転çïŒå転ïŒãåé¢ã«åºãŸãã ãããã®è©äŸ¡åºæºã®ãããããä»ã«åããã®ã¯ãããŸããã ãã¬ã€ã€ãŒã圌ãã®åœ±é¿ã«é©å¿ããã«ã€ããŠãåžå Žç¶æ³ãžã®åœ±é¿ã¯å€ãããŸãã å¹çæ§ä»®èª¬ã§ã¯ããããããã®ãåžå Žã®ç°åžžãšåŒã³ãŸãããããã¯å®å šã«çå®ã§ã¯ãããŸããã ãããã®å€èŠ³ã¯éåžžã«äºæž¬å¯èœã§ãããåžå Žãžã®åœ±é¿ãå®éåã§ããŸãã
æ¹å€ã«å¿ããŠãåžå Žå¹çã®æ¯æè ã¯2ã€ã®æ°ãã仮説ãç«ãŠãŸããã å ¬éæ å ±ã®ã¿ãåŒçšç¬Šã«åæ ãããŠãããšäž»åŒµãã人ãããŸãã ãã1ã€ã¯ãçŸåšã®æ ªäŸ¡ã«å¯Ÿããå±¥æŽããŒã¿ã®åœ±é¿ãèªèããããšã§ãã ããããäž»ãªäœçœ®ã¯åããŸãŸã§ããäŸ¡æ Œã®åãã¯ããã«ã³ãé£éãã©ã³ãã ãŠã©ãŒã¯ã®ååžã«åŸããŸãã
ã»ãŒãã¹ãŠã®çŸä»£ã®äŸ¡æ Œã¢ãã«ã®æ ¹åºã«ããã©ã³ãã ãŠã©ãŒã¯ä»®èª¬ãè€è£œããŠããããšãããããŸãã åŠè ãšåžå Žã®å®è·µè ã¯ãåžå Žã®çžå¯Ÿçãªæå¹æ§ãšã©ã³ãã æ§ãç 究ããããã«æ°å幎ãè²»ãããŠããŸãã åã®èšäºã§ãStuart Reedã¯ãã©ã³ãã æ§ããã§ãã¯ãããã¹ãã®1ã€ãLoããã³McKinleyã®äžåäžåæ£é¢ä¿ãã¹ãã玹ä»ããŸããã ä»åã¯ããšããããLempel-Zivå§çž®ã¢ã«ãŽãªãºã ã«çŠç¹ãåœãŠãŸãã
ã¢ã«ãŽãªãºã æ å ±çè«ãšã©ã³ãã ãã¹
å®éåæã§ã¯ãã©ã³ãã æ§ã®ç 究ã¯æ°ãããã®ã§ã¯ãããŸããã æå·åãéåååŠãããã³éºäŒåŠã¯ãããé·ãéè¡ã£ãŠããŸããã åœç¶ã®ããšãªããããããã®ç§åŠããããã€ãã®ã¢ã€ãã¢ãã¢ãããŒããåãå ¥ããããŸããã ããšãã°ããšã³ããã¹ãã¯ãæ å ±ã®ã¢ã«ãŽãªãºã çè«ãããã©ã³ãã ã·ãŒã±ã³ã¹ã¯éå§çž®æ§ã§ãããšããèããåãå ¥ããŠããŸãã ãŸãããã®èå³æ·±ãèãã®èæ¯ã確ç«ããŠã¿ãŸãããã
1930幎代ãAlan Turingã¯èªåèšç®æ©ãšåŒã°ããæœè±¡çãªèšç®ã¢ãã«ãå°å ¥ããŸããã ä»æ¥ã§ã¯ããã¥ãŒãªã³ã°ãã·ã³ãšããŠç¥ãããŠããŸãã ãã®ç¹ã§ãç§ãã¡ã¯æ¬¡ã®å£°æã«ã®ã¿èå³ããããŸãããã©ã®èšç®ã¢ã«ãŽãªãºã ãå®ããããŠãããã¯é¢ä¿ãããŸããããã·ã³ã¯ã¢ã«ãŽãªãºã èªäœã®ããžãã¯ãæš¡å£ããå¿ èŠããããŸããã
ãã¥ãŒãªã³ã°ãã·ã³ã¢ãã«
ç°¡ç¥åããã圢åŒã§ã¯ãã¢ã«ãŽãªãºã ã¯å ¥åããŒã¿ã®ã»ãããåºåã®ã»ããã«å€æããããã®æ±ºå®è«çãªäžé£ã®ã¹ãããã§ãã ãã£ããããæ°åã«å¯ŸããŠãã®ãããªã¢ã«ãŽãªãºã ãæ§ç¯ããã®ã¯ç°¡åã§ãã次ã®æ°ã¯åã®2ã€ã®æ°ã®åèšã§ãã å¥ã®å倧ãªæ°åŠè ã¢ã³ãã¬ã€ã»ã³ã«ã¢ãŽããã«ãããšãã·ãŒã±ã³ã¹ã®è€éãã¯ããã®ã·ãŒã±ã³ã¹ãåºåãšããŠçæããæçã®èšç®ããã°ã©ã ã®é·ãã«ãã£ãŠæž¬å®ã§ããŸãã ãã®ããã°ã©ã ãå ã®ã·ãŒã±ã³ã¹ãããçãå Žåã¯ãã·ãŒã±ã³ã¹ãå§çž®ãã€ãŸããšã³ã³ãŒãããŸãã
ãã¥ãŒãªã³ã°ãã·ã³ã«ã©ã³ãã ã·ãŒã±ã³ã¹Sãå ¥ãããšã©ããªããŸããïŒ ããŒã¿ã®é åžã«ããžãã¯ããªãå Žåããã·ã³ã«ã¯ã·ãã¥ã¬ãŒããããã®ããããŸããã ã©ã³ãã ã·ãŒã±ã³ã¹ãååŸããå¯äžã®æ¹æ³ã¯print Sã§ãããã ãããã®ããã°ã©ã ã¯ã©ã³ãã ã·ãŒã±ã³ã¹Sãããé·ããªããŸãããããã£ãŠããã®è€éãã¯å®çŸ©ã«ããé«ããªããŸãã ããã¯ãã©ã³ãã ã·ãŒã±ã³ã¹ãéå§çž®æ§ã§ãããšèšããšãã®æå³ã§ãã å¥ã®è³ªåãèŠã€ããããã°ã©ã ãæçã§ãããšããäºå®ãã©ã®ããã«ç¢ºèªã§ããŸããïŒ ãã®çµè«ã«å°éããããã«ãç¡éã®çµã¿åãããæŽçããå¿ èŠããããŸãã
ç°¡åã«èšãã°ãã³ã«ã¢ãŽããã®è€éãã¯èšç®ã§ããŸããã ãããã£ãŠãã©ã³ãã æ§ã¯èšç®ã§ããŸããã ã·ãŒã±ã³ã¹ãã©ã³ãã ã§ããããšã蚌æããããšã¯ã§ããŸããã ããããããã¯ãã·ãŒã±ã³ã¹ãã©ã³ãã ã§ããå¯èœæ§ããã¹ãã§ããªããšããæå³ã§ã¯ãããŸããã ããã§ãçµ±èšçã©ã³ãã æ§ãã¹ããç»å Žãããã®äžéšã¯å§çž®ã¢ã«ãŽãªãºã ã䜿çšããŠããŸãã
å§çž®ãšåžå Žã®ã©ã³ãã æ§ïŒåé¡ã®èª¬æ
åžå Žã®ããã©ãŒãã³ã¹ããã¹ãããããã«å§çž®ã¢ã«ãŽãªãºã ãé©åãããããšããç§åŠè ã¯ãããã§ãã ã¢ã€ãã¢èªäœã¯åçŽã§ããããã®åçŽãã®èåŸã«ç¹å®ã®èª²é¡ãé ãããŠããŸãã
課é¡1ïŒæéã·ãŒã±ã³ã¹ãšç¡é倧
æ å ±çè«ã«ãããã©ã³ãã æ§ã®å®çŸ©ã¯ãç¡éã®ç¯å²ã«å¯ŸåŠããå¿ èŠãããããšãæå³ããŸãã 確çã®ã¿ã«åºã¥ããŠãä»»æã®æéã·ãŒã±ã³ã¹ããã§ã«ãããã«å§çž®ã§ããŸãã åžå Žã§æ¯èŒççãã·ãŒã±ã³ã¹ããã§ãã¯ããããã«å§çž®ã䜿çšããå Žåãã¢ã«ãŽãªãºã ãããããå§çž®ã§ãããã©ããã¯æ°ã«ããŸããã å§çž®çãçµ±èšçã«ææã«ãªããã©ããçåã«æã£ãŠããŸãã
課é¡2ïŒåžå Žã®ããªãã
ã©ã³ãã æ§ã«ã¯ããã€ãã®çš®é¡ããããŸãã æåã®ã¿ã€ãã¯ããã«ãã³ã²ãŒã«ãŸãã¯ã³ã«ã¢ãŽããã®æ¬åœã®äºæ ãšåŒã°ããŸãã ã»ãšãã©ã®å Žåãã³ã³ãã¥ãŒã¿ãŒãµã€ãšã³ã¹ãšæ å ±çè«ã«é¢é£ããŠããŸãã 2çªç®ã®ã¿ã€ãã¯ãéèåæããã³ã©ã³ãã ãŠã©ãŒã¯ä»®èª¬ã§äœ¿çšããããã«ã³ãã©ã³ãã ãã¹ã§ãã
ããã¯ãåžå Žã®ã©ã³ãã æ§ã®çµ±èšçæ€å®ãéžæããéã«ã¯éåžžã«æ³šæããªããã°ãªããªãããšãæå³ããŸãã æåã®ã¿ã€ãã®ã©ã³ãã æ§ã®ããã«èšèšãããçµ±èšçæ€å®ã¯ããã«ã³ãã©ã³ãã æ€å®ãšã¯ç°ãªãä¿¡é Œåºéãæã¡ãŸãã
ããã§å°ãã¯ãªã¢ã«ãªããŸãã ãã©ãŠã³éåãšéãŒãåå·®ã®å¹ŸäœåŠçãã©ãŠã³éåã®2ã€ã®ã¢ãã«ããããšããŸãã ãã®äŸã§ã¯ãæåã®ã¢ãã«ã¯äž¡æ¹ã®ã¿ã€ãã®ã©ã³ãã æ§ã«é©ããŠããŸãã 2çªç®ã¯ããã«ã³ãã®ãã£ã³ã¹ã®ããã ãã§ãã 2çªç®ã®ã¢ãã«ã®ãã¹ãçµæã¯æªããããŸãã
ã¢ãã«1ïŒãã©ãŠã³éåïŒ-å¹³åããŠããã¹ãŠã®çµè·¯ã¯ã©ãã«ãå°ããããåå·®ã¯ãŒãã§ãã äºå€åã«ããããã®ã¢ãã«ã¯ã³ã€ã³ããªããã·ãŒã±ã³ã¹ã«å€ãããŸã
ã¢ãã«2ïŒå¹ŸäœåŠçãªãã©ãŠã³éåïŒ-çµè·¯ãé²ã¿ãåå·®ã¯ãŒãã§ã¯ãããŸããã äºå€åäžã«ãçµæã«æªã¿ãå°å ¥ãããŸã
ããã¯ãã©ã³ãã ãŠã©ãŒã¯ã¹ã«ãŒå§çž®ã®ä»®èª¬ã«åè«ããã人ã«ãšã£ãŠæ·±å»ãªåé¡ã§ãã çè«ã¯ãåžå Žã®ããªããã®é床ã®äºæž¬å¯èœæ§ãã€ãŸãã誰ããåžå Žãç Žãèœåã®ã¿ãæ±ã£ãŠããããã§ãã
課é¡3ïŒç¢ºççãã©ãã£ãªãã£ãšã®ã£ãã
次ã®åé¡ã¯ããã©ãã£ãªãã£ã®äžå®å®æ§ããç¹å®ã®åžå Žã®åããçè«çèšç®ã«åããŠãèŠåçã§é »ç¹ãªçŸè±¡ã®ããã«èŠãããšããäºå®ã«ã€ãªããå¯èœæ§ãããããšã§ãã ãã®åé¡ã¯ãåãå°ŸããšãåŒã°ããŸãã ã®ã£ããã確ççãã©ãã£ãªãã£ã«é¢ä¿ãªããåžå Žã®å€åã®ãã€ãºããŒãã«ãªãåŸåããããšä»®å®ãããšããã€ããªã³ãŒãã®å€åãæ³åããŠåé¡ã解決ã§ããŸãã
åžå Žã®ã©ã³ãã æ§ã®å§çž®ãã¹ã
ã¢ã«ãããã¯ãµã€ãšã³ã¹ã§ã¯ããã¹ã¬ã¹ããŒã¿å§çž®ã¢ã«ãŽãªãºã ãåžå Žã®ããã©ãŒãã³ã¹ãè©äŸ¡ããããã«ãã§ã«é©çšãããŠããŸãã ã»ãšãã©ã®ç 究ã®æ¬è³ªã¯æ¬¡ã®ãšããã§ããå é²åžå Žã¯é«ã¬ãã«ã®å¹çæ§ã瀺ããæ°èåžå Žã¯é床ã®å§çž®æ§ãšäžå®ã¬ãã«ã®éå¹çæ§ã瀺ããŸãã ãããã®éå¹æã¯ãç¹å®ã®æéééã§äºæ³ããããããé«ã確çã®ã¬ãã«ã瀺ãç¹å®ã®ãã¿ãŒã³ã«ãã£ãŠã®ã¿èª¬æã§ããŸãã
æ€èšŒæ¹æ³
æ¹æ³è«ã¯ã3ã€ã®ããç¥ãããå§çž®ã¢ã«ãŽãªãºã ã«åºã¥ããŠããŸãïŒ{gzipãbzip2ããã³xz}ã
- Quandl.comããããŒã±ããã€ã³ããã¯ã¹ãããŠã³ããŒãããå¿ èŠããããŸãã
- 次ã«ãã€ã³ããã¯ã¹ã®å¯Ÿæ°ãªã¿ãŒã³ãèšç®ããrãšããŠæå®ããŸãã
- ãããã®å¹³åïŒåå·®æåïŒ-ÎŒrãèšç®ããŸãã
- æåæ§ïŒãã¬ã³ãïŒãé€å»ããã«ã¯ãrÎŒrãæžç®ããŸãã
- 次ã«ãrã¯ãã€ããªã·ã¹ãã ã«å€æããããŒããã倧ããå€ã1ã§çœ®ãæãããŒãããå°ããå€ããããããŒãã§çœ®ãæããŸãã
- 次ã«ãrã¯ãåå¥ã®ãŠã£ã³ããŠã®ã€ã³ããžã·ã§ã³ã«ãã£ãŠmã«åå²ãããŸã-Wã¯ãããã7幎éã§ãã
- åãŠã£ã³ããŠã¯16é²æ°ã®whïŒ4æ¥éã®ãµãã·ãŒã±ã³ã¹ïŒã«å€æãããŸãã
- åw h ãw hcã¯å§çž®ãããå§çž®ä¿æ°ãèšç®ãããŸã ã
- ãã®å¹³åã¯ããã¹ãŠã®ãŠã£ã³ããŠ-c âã«åæ£ãããŸãã
- äºæ³ãããå§çž®çã¯ãæ¬äŒŒã©ã³ãã ããŒã¿-E [c â]ã䜿çšããŠèšç®ãããŸãã
- c â > = minïŒ1.0ãE [c â ]ïŒã®å Žåãåžå Žã¯é床ã®å§çž®æ§ã瀺ãããããã¯ãã®æå¹æ§ãæå³ããŸãã
- c â <minïŒ1.0ãE [c â ]ïŒã®å Žåãé床ã®å§çž®çãååšããåžå Žã®éå¹çæ§ãæå³ããŸãã
Rãã¹ãããã³è§£æã¢ã«ãŽãªãºã
Rã«ã¯memCompressããã³memDecompressé¢æ°ãæ¢ã«ããããããã®ãããªãã¹ããè¡ãã®ã«ããã»ã©å€ãã®ã³ãŒãã¯å¿ èŠãããŸããã§ããã Pythonã«ãåæ§ã®æ©èœããããŸãã
compressionTest <- function(code, years = 7, algo = "g") { # The generic Quandl API key for TuringFinance. Quandl.api_key("t6Rn1d5N1W6Qt4jJq_zC") # Download the raw price data. data <- Quandl(code, rows = -1, type = "xts") # Extract the variable we are interested in. ix.ac <- which(colnames(data) == "Adjusted Close") if (length(ix.ac) == 0) ix.ac <- which(colnames(data) == "Close") ix.rate <- which(colnames(data) == "Rate") closes <- data[ ,max(ix.ac, ix.rate)] # Get the month endpoints. monthends <- endpoints(closes) monthends <- monthends[2:length(monthends) - 1] # Observed compression ratios. cratios <- c() for (t in ((12 * years) + 1):length(monthends)) { # Extract a window of length equal to years. window <- closes[monthends[t - (12 * years)]:monthends[t]] # Compute detrended log returns. returns <- Return.calculate(window, method = "log") returns <- na.omit(returns) - mean(returns, na.rm = T) # Binarize the returns. returns[returns < 0] <- 0 returns[returns > 0] <- 1 # Convert into raw hexadecimal. hexrets <- bin2rawhex(returns) # Compute the compression ratio cratios <- c(cratios, length(memCompress(hexrets)) / length(hexrets)) } # Expected compression ratios. ecratios <- c() for (i in 1:length(cratios)) { # Generate some benchmark returns. returns <- rnorm(252 * years) # Binarize the returns. returns[returns < 0] <- 0 returns[returns > 0] <- 1 # Convert into raw hexadecimal. hexrets <- bin2rawhex(returns) # Compute the compression ratio ecratios <- c(ecratios, length(memCompress(hexrets)) / length(hexrets)) } if (mean(cratios) >= min(1.0, mean(ecratios))) { print(paste("Dataset:", code, "is not compressible { c =", mean(cratios), "} --> efficient.")) } else { print(paste("Dataset:", code, "is compressible { c =", mean(cratios), "} --> inefficient.")) } } bin2rawhex <- function(bindata) { bindata <- as.numeric(as.vector(bindata)) lbindata <- split(bindata, ceiling(seq_along(bindata)/4)) hexdata <- as.vector(unlist(mclapply(lbindata, bin2hex))) hexdata <- paste(hexdata, sep = "", collapse = "") hexdata <- substring(hexdata, seq(1, nchar(hexdata), 2), seq(2, nchar(hexdata), 2)) return(as.raw(as.hexmode(hexdata))) }
Githubã³ãŒã
以äžã¯ãå§çž®ã¢ã«ãŽãªãºã ã§äœ¿çšããã3ã€ã®å§çž®ã¹ããŒã ã®ç°¡åãªèª¬æã§ãã
- Gzipå§çž® LZ1ãšãããã³ã®ãšã³ã³ãŒããçµã¿åãããDEFLATEã¢ã«ãŽãªãºã ã«åºã¥ããŠããŸãã LZ1ã¯ãã¢ãã©ãã ã¬ã³ãã«ãšãžã§ã€ã³ããžãã«ã¡ãªãã§åœåãããŸããã ã·ãŒã±ã³ã¹å ã®ç¹°ãè¿ããã¿ãŒã³ã眮ãæãããããã®ãã¿ãŒã³ã®ã³ããŒãžã®ãªã³ã¯ãå ã®éå§çž®ããŒãžã§ã³ã«é 眮ããããšã§æ©èœããŸãã åäžèŽã¯ããã³ãŒããŒã«é¢ãã次ã®æ å ±ãå«ãé·ããã¢ãä»ããŠãšã³ã³ãŒããããŸãããã©ã¡ãŒã¿ãŒyã¯ããã®åŸã®è·é¢xã®ãã¹ãŠã®ãã©ã¡ãŒã¿ãŒãšçãããªããŸãã David Huffmanã«ã¡ãªãã§åä»ãããã2çªç®ã®ã¢ã«ãŽãªãºã ã¯ãã·ãŒã±ã³ã¹å ã®ãã¿ãŒã³ã®åºçŸé »åºŠã«ãã£ãŠæé©ãªãã¬ãã£ãã¯ã¹ããªãŒãæ§ç¯ããŸãã
- BZIP2å§çž® ã Burroughs-Wheelerå€æãmotion-to-startå€æïŒMTFïŒãããã³Huffmanã³ãŒãã£ã³ã°ã®3ã€ã®å§çž®ã¡ã«ããºã ã䜿çšããŸãã ãããã®æåã®æ¹æ³ã¯ãã·ãŒã±ã³ã¹ãé¡äŒŒã®æåã®ã¹ããªãŒã ã«åé 眮ããŠãæšæºçãªæ¹æ³ã§å§çž®æé ãç°¡çŽ åããå¯éçãªæ¹æ³ã§ãã MTFå€æã¯ãã·ãŒã±ã³ã¹å ã®æåãã¹ã¿ãã¯äžã®ã€ã³ããã¯ã¹ã«çœ®ãæããŸãã ãããã£ãŠãåãæåã®é·ãã·ãŒã±ã³ã¹ã¯å°ããæ°åã«çœ®ãæãããããŸããªæåã¯å€§ããªæåã«çœ®ãæããããŸãã
- XZå§çž® ã ããã¯ã7zå§çž®åœ¢åŒã§éçºãããäžé£ã®Lempel-Ziv-Markovã¢ã«ãŽãªãºã ãéããŠæ©èœããŸãã ããã¯åºæ¬çã«ãLZ1ã¢ã«ãŽãªãºã ã«ç¶ãèªåœå§çž®ã¢ã«ãŽãªãºã ã§ãã ãã®åºåã¯ãé©å¿ãã€ããªã¬ã³ãžãšã³ã³ãŒããŒãšåŒã°ãããåãããã®ç¢ºççäºæž¬ãæäŸããã¢ãã«ã䜿çšããŠãšã³ã³ãŒããããŸãã
ããšãã°ã3ã€ã®ã¢ã«ãŽãªãºã ã¯ãã¹ãŠãé¢æ°memCompressããã³memDecompressãä»ããŠRèšèªã§äœ¿çšã§ããŸãã ãããã®é¢æ°ã®å¯äžã®åé¡ã¯ãå ¥åã16é²ã·ãŒã±ã³ã¹ã®åœ¢åŒã§äžããããªããã°ãªããªãããšã§ãã bin2rawhexé¢æ°ã䜿çšããŠå€æã§ããŸãã
çµæãšãã®oïŒæªéïŒã®æå³
å®éšã§ã¯ã51ã®ã°ããŒãã«åžå Žã€ã³ããã¯ã¹ãš12ã®é貚ãã¢ã䜿çšãããŸããã ã·ãŒã±ã³ã¹ã¯ãæ¯é±ããã³æ¯æ¥ã®é »åºŠã§ãã¹ããããŸããã
ãã¹ãŠã®ã¢ã»ããã¯ããŠã©ãŒã¯ãã©ã¯ãŒãææ³ã䜿çšããgzipãbzip2ãããã³xzã¢ã«ãŽãªãºã ã䜿çšããŠå§çž®ãããŸããã ãã«/ã«ãŒãã«ã®ãã¢ãé€ããã©ã®è³ç£ãé床ã®å§çž®æ§ã瀺ããŸããã§ããã å®éããããã®ã©ããå§çž®æ§ããŸã£ãã瀺ããŸããã§ããã ãããã®çµæã¯ãç·åœ¢è€é床ãã¹ãã䜿çšããéå»ã®å®éšã®çµæãè£ä»ããŸããã ããããNISTãã¹ãã®çµæãšLo-McKinleyãã¹ãã®çµæã¯ççŸããŠããŸãã çç±ãç解ããå¿ èŠããããŸãã
ãªãã·ã§ã³1ãæåã®çµè«ã¯ãã°ããŒãã«åžå Žãæ¯èŒçå¹ççã§ããããšãïŒåŠè ãåã°ããããã«ïŒèªèããããšã§ãã ãããã®åäœã¯éåžžã«ã©ã³ãã ã«èŠããŸãã
ãªãã·ã§ã³2ããŸãã¯ãéèåžå Žã®å¹çãšã©ã³ãã æ§ã®ãã§ãã¯ã«é¢é£ããŠãããŒã¿æ倱ã®ãªãå§çž®ã¢ã«ãŽãªãºã ã«ééããããããšãèªããããšãã§ããŸãã
ãããã®ä»®èª¬ããã¹ãããããã«ãå®éšã®èè ã¯å§çž®ãã¹ãããã¹ããããã¹ããéçºããå¿ èŠããããŸããã
å§çž®è©Šéš
å®éãã©ã³ãã ãŠã©ãŒã¯ä»®èª¬ã®å察è ã¯ã100ïŒ ã®åžå Žæ±ºå®è«ã䞻匵ããŠããŸããã ãã€ã³ãã¯ãåžå Žã®è¡åã100ïŒ ã©ã³ãã ã§ã¯ãªãããšã蚌æããããšã§ãã ã€ãŸããåžå Žã«ã¯ã·ã°ãã«ãšãã€ãºã®2ã€ã®åŽé¢ããããŸãã ä¿¡å·ã¯äœçŸãã®ç°åžžãšããŠçŸãããã€ãºã¯æãããªäžèŽãšå¹çãšããŠæ©èœããŸãã ãããã¯ã³ã€ã³ã®äž¡é¢ã§ãã 次ã®ã¹ãããã¯ãã·ã¹ãã ã«ãã£ãŠçæããããã€ãºãä¿¡å·ãçžæ®ºããªãããšãèªèããããšã§ãã ãã®ãããªä¿¡å·ãèŠã€ããŠäœ¿çšããããšã«ãããåžå Žãããããã³ã°ãããæ©äŒãæ®ã£ãŠããŸãã
åè¿°ã«åºã¥ããŠãèè ã¯å§çž®ãã¹ãããã¹ãããããã«æ¬¡ã®åŒãæäŸããŸãã
ãã€ãºæåã®æšæºåå·®--tã®ãµã€ãºã倧ãããããšãåžå Žã®å§çž®æ§ãã©ã®ããã«æªåãå§ããã®ããããããŸãã æšæºåå·®ã0.01ããŒã¯ãè¶ ãããšãå§çž®ãã¹ãã¯å€±æãå§ããŸãã
Ï= 0.010ãÏ= 0.020ããã³Ï= 0.040ã®ãã¹ãã®çµè«ã®äŸã¯æ¬¡ã®ãšããã§ãã
å§çž®ãã¹ãã§ã¯ããããã®åžå Žã¯ãã¹ãŠããŒã¿æ倱ãªãã«å§çž®ã¢ã«ãŽãªãºã ã§å§çž®ã§ããªããããã©ã³ãã ãŸãã¯å¹ççã§ãã ã¯ãããããã¯ã©ã³ãã ã«èŠãããããããŸããã ãããããããã¯ä¿¡å·ãšãã€ãºã§æ§æãããŠãããããã©ã³ãã æ§ãå®å šã«ããããšã¯ã§ããŸããã
ããé©åãªãã§ãã¯ãè¡ãã«ã¯ãå¹³åã«æ»ãããã«ãããæŠç¥/ã¢ã«ãŽãªãºã ã䜿çšããå¿ èŠããããŸãã
- çµè«ïŒãŠã£ã³ããŠãµã€ãºw = 4;
- æéãã©ã¡ãŒã¿tã«ã€ããŠãéå»wãã€ãã¹æ¥t-cãžã®çŽ¯ç©ãªã¿ãŒã³ãèšç®ããŸãã
- <0,0ã®å Žåãè³ç£ã®100ïŒ ãä¿æããŸãïŒè³ç£ãäžãããšç¢ºä¿¡ããŠããŸãïŒã
- ãã®ä»ã®å Žåã¯ãå ±æãåé€ããŸãã
以äžã¯ãåãã€ãºã¬ãã«ã®2ã€ã®å³ã§ãã æåã®ãã®ã¯ãå¹³åã¬ãŒãæŠç¥ãžã®ååž°ã«ãã£ãŠçæããã30ã®æ ªäŸ¡æ²ç·ã瀺ããŠããŸãã 2çªç®ã®äŸã¯ãæŠç¥ã®é©çšã«ããåå ¥ãšæ¯èŒãããåã®äŸã®å¹³ååžå Žåå ¥ã瀺ããŠããŸãã
ããã¯ç°¡åãªäŸã§ãããççŸãç¹å®ããã®ã«åœ¹ç«ã¡ãŸãã
åžå Žãå§è¿«ãããŠããªãå Žåãããã¯å¹æçã§ãããšããããšã«ãªããŸããããããæ¬åœã«å¹æçã§ããå Žåããããããããã³ã°ãããè³ãæŠç¥ãé©çšããããšã¯äžå¯èœã§ãã
ã©ã³ãã æ§ã¯èšŒæã§ããªããããã·ãŒã±ã³ã¹ã¯å®éã«ã¯ã©ã³ãã ã§ã¯ãªããããããããç解ããŠèš±ããå¿ èŠããããŸãã èè ã«ãããšãå§çž®ãã¹ãã¯ååŒæŠç¥ããããã€ãºã«ææã§ãããã倱æããŸããïŒã·ãã¥ã¬ãŒã·ã§ã³ã§ã¯ããã€ãºã«å¯ŸããŠ10åã®æ床ãåŸãããŸãïŒã çç±ã¯ãããŒã¿æ倱ã®ãªãå§çž®ã¢ã«ãŽãªãºã ã®äœ¿çšãããããŸããã
æ€èšŒãã¹ãã®ã³ãŒãã¯ãã¡ãã«ãããŸã ã
çµè«
ãã®ç 究ã¯ãå§çž®ã¢ã«ãŽãªãºã ã䜿çšããŠåžå Žã®ã©ã³ãã æ§ããã¹ãããæ¹æ³ã®ã¢ã€ãã¢ãæ瀺ãããããåžå Žã®æå¹æ§ãç解ããããšã瀺ããŠããŸãã 次ã«ããã¡ã¿ãã¹ãããéå§ãããŸãããããã¯ããã€ãºã«å¯Ÿãããã®ãããªãã¹ãã®æ床ãé«ãããšã瀺ããŠããŸãã ãã®çµæãããã€ãã®çµè«ãåºãããšãã§ããŸãã
- ãã¬ã³ãé€å»åŸãåžå Žã¯ååãªãã€ãºã瀺ãå§ããããŒã¿æ倱ã®ãªãå§çž®ã¢ã«ãŽãªãºã ïŒgzipãbzip2ãxzïŒã«å¯ŸããŠãééäžèœãã«ãªããŸãã
- ãã€ãºã«å¯Ÿãããã¹ãã®æ床ãååŒæŠç¥ã®æ床ãããé«ãå ŽåïŒèšäºã®äŸã§ã¯ããã®ãããªãã¹ãã¯åçŽãªè³ãæŠç¥ããããã€ãºã«å¯ŸããŠ10åææã§ããããšãå€æããŸããïŒãåŠå®çãªçµæã¯åæãããåžå Žãå¹æçã§ããããšãæå³ããŸããã ããããã³ã°ãã§ããªããšããæå³ã§ã
- èšãæããã°ããã®ãããªãã¹ãã¯ãã©ã³ãã ãªåžå Žã®anderã®ä»®èª¬ã®å³å¯ãªè§£éã䞻匵ããåŠè ã®æ£ããã蚌æããªãã äžèŠãããšãããããè€éã§ããããšãå€æããŸããã
- åžå Žã®ã©ã³ãã æ§ã決å®ããçµ±èšçãã¹ãã¯ããµã€ãºããªãŒããããã³æ床ãç°ãªããŸãã ä¿¡é Œã§ããçµæãåŸãããã®ã¯ãããã€ãã®ãã¹ãã®çµã¿åããã®ã¿ã§ãã
ITinvestã®ãã®ä»ã®é¢é£è³æïŒ
- ITinvestã®å°é家ã«ããéèåžå Žã®ç¶æ³ã«é¢ããåæè³æ
- ã€ã³ã¿ãŒãããäžã®ã¡ãã»ãŒãžã®ææ ãåæããããšã«ãããååŒæã§ååŒããããã®ã¢ã«ãŽãªãºã ãäœæããããšã¯å¯èœã§ããïŒ
- æ ªåŒåžå ŽååŒã«æé©ãªæéã決å®ããæ¹æ³ïŒã¢ã«ãŽãªãºã ã«åŸãåŸå
- IBã®ã¹ã¿ãŒãã¢ããPalantirã¯ãæè³éè¡Credit Suisseãäžæ£ãªãã¬ãŒããŒãèå¥ããããã®ã·ã¹ãã ã®äœæãæ¯æŽããŸã
- å®éšïŒGoogleãã¬ã³ãã䜿çšããŠæ ªåŒåžå Žã®ã¯ã©ãã·ã¥ãäºæž¬ãã
- å®éšïŒæ ªäŸ¡ææ°ã®æ¯ãèããäºæž¬ããã¢ã«ãŽãªãºã ã®äœæ