ngµç×ÓÓÎÏ·

05

¿ÆÑÐÏ£Íû

Ñо¿Ìá³ö»ùÓÚѸËÙ¶ÈÐÅÏ¢µÄCVaR¶¯Ì¬ÓÅ»¯ÀíÂÛÓëËã·¨

¸å¼þȪԴ £º£º£ºÖÎÀíѧԺ ±à¼­ £º£º£ºËïè¡¡¢Íõ¶¬Ã· ÉóºË £º£º£ºËïÒ«±ó ÔĶÁÁ¿ £º£º£º

ngµç×ÓÓÎÏ·ÐÂÎÅÍøÑ¶£¨Í¨Ñ¶Ô±ÏÄÀþ£©½üÆÚ£¬£¬£¬ngµç×ÓÓÎÏ·ÖÎÀíѧԺÏÄÀþ½ÌÊÚÔÚÖÎÀíѧÁìÓò¹ú¼Ê¸ßˮƽÆÚ¿¯Production and Operations Management£¨¼ò³ÆPOM£©ÉϽÒÏþÁËÌâΪ¡°Risk-Sensitive Markov Decision Processes with Long-Run CVaR Criterion¡±µÄÑо¿ÂÛÎÄ£¬£¬£¬ÂÛÎĵįäËû×÷Õß»¹°üÀ¨ngµç×ÓÓÎÏ·ÖÎÀíѧԺµÄ²©Ê¿ÉúÕÅè´ÑþºÍ˹̹¸£´óѧÖÎÀí¿ÆÑ§Ó빤³ÌϵµÄPeter W. Glynn ½ÌÊÚ¡£¸ÃÑо¿Õë¶ÔËæÎÞа̬ϵͳÖеÄÀú³ÌÖÐËðʧµÄCVaRÓÅ»¯ÎÊÌâ¾ÙÐÐÑо¿£¬£¬£¬ÍêÉÆÁËÏìÓ¦µÄÓÅ»¯ÀíÂÛ¼°Ë㷨ϵͳ¡£

CVaRÖ¸±êÊÇÖ÷ÒªµÄ·çÏÕÃè»æÖ¸±ê£¬£¬£¬ÔÚÓ¦ÓÃÓÚ¶à½×¶Î¶¯Ì¬¾öÒéʱ£¬£¬£¬ÓÉÓÚÖ¸±êº¯ÊýµÄ²»¿É¼ÓÐÔµ¼Ö¾­µä¶¯Ì¬ÍýÏëÔ­ÀíʧЧ£¬£¬£¬Bellman×îÓÅÐÔ·½³Ì²»¿ÉÁ¢£¬£¬£¬ÐèҪ׷ÇóеÄÓÅ»¯ÒªÁì¡£±¾ÎÄ»ùÓÚѸËÙ¶ÈÓÅ»¯ÒªÁì¶ÔÀëɢʱ¼äÎÞÏÞ½×¶ÎÎÈ̬CVaR ×¼ÔòϵÄÂíÊϾöÒéÀú³Ì£¨MDP£©ÓÅ»¯ÎÊÌâ¾ÙÐÐÑо¿¡£Í¨¹ýÒýÈëα CVaR Ö¸±ê£¬£¬£¬½«Ô­ÎÊÌâת»¯ÎªÒ»¸öÁ½²ãMDPÎÊÌ⣬£¬£¬ÄÚ²ãΪ±ê×¼¶¯Ì¬ÍýÏëÎÊÌ⣬£¬£¬Íâ²ãΪαCVaRµÄµ¥²ÎÊýÓÅ»¯ÎÊÌ⣬£¬£¬²¢¸ø³öÁË CVaRÐÔÄܲî·Ö¹«Ê½ÓÃÒÔÃè»æ²î±ðÕ½ÂÔ¶ÔÓ¦µÄÎÈ̬ CVaR ÐÔÄܲî¡£

ÂÛÎÄ֤ʵÎúÈ·¶¨ÐÔÆ½ÎÈÕ½ÂÔµÄ×îÓÅÐÔ£¬£¬£¬»ùÓÚCVaR²î·Ö¹«Ê½ºÍÐÔÄܵ¼Êý¹«Ê½»ñµÃÁËCVaR Bellman¾Ö²¿×îÓÅ·½³Ì£¬£¬£¬´Ó¶ø¸ø³öÁË»ñµÃ¾Ö²¿×îÓÅÕ½ÂԵijäÒªÌõ¼þÒÔ¼°ÎÈ̬CVaR MDPµÄÕ½ÂÔµü´úÐÍËã·¨£¬£¬£¬Ö¤ÊµÎú¸ÃËã·¨¿ÉÊÕÁ²ÖÁ¾Ö²¿×îÓÅÕ½ÂÔ¡£½øÒ»²½£¬£¬£¬ÂÛÎÄ»ùÓÚÁ½²ãMDPÎÊÌâµÄѸËÙ¶ÈÐÅÏ¢ºÍÁÙ½çµãÆÊÎö£¬£¬£¬Ö¤ÊµÎúαCVaRº¯ÊýµÄ·ÖƬÏßÐÔ¡¢·Ö¶Î͹µÄÐÔ×Ó£¬£¬£¬ÔÚ´Ë»ù´¡Éϸø³öÁËÒ»ÖÖÈ«¾Ö×îÓÅËã·¨£¬£¬£¬Ö¤ÊµÎúËã·¨¿ÉÊÕÁ²ÖÁÈ«¾Ö×îÓÅÕ½ÂÔ¡£ÂÛÎÄ×îºóͨ¹ý¶à¸öÊýֵʵÑé±ÈÕÕÑéÖ¤Á˱¾ÎÄÓÅ»¯ÀíÂÛÓëËã·¨µÄÓÐÓÃÐÔ¡£

ÂÛÎĵÄÖ÷ҪТ˳¿É·ÖΪÒÔÏÂÈýµã£¬£¬£¬µÚÒ»£¬£¬£¬±¾ÎÄÊ״ζÔȨºâϵͳÀú³Ì²¨¶¯ÐÔµÄÎÈ̬CVaR×¼ÔòϵÄMDPÓÅ»¯ÀíÂÛ¾ÙÐÐÑо¿£¬£¬£¬ÍêÉÆÁËÏÖÓÐÎÄÏ×ÔÚ¸ÃÀàÖ¸±êµÄÀíÂÛϵͳ£»µÚ¶þ£¬£¬£¬²î±ðÓÚ¾­µäMDPÀíÂÛ£¬£¬£¬±¾ÎÄ´ÓѸËÙ¶ÈÓÅ»¯µÄ½Ç¶È¶ÔÎÈ̬CVaR MDP¾ÙÐÐÑо¿£¬£¬£¬»ñµÃÁËCVaR ÐÔÄܲî·Ö¹«Ê½¡¢ÐÔÄܵ¼Êý¹«Ê½ÒÔ¼° CVaR Bellman ¾Ö²¿×îÓÅ·½³Ì£»µÚÈý£¬£¬£¬Í¨¹ý½«Ô­ÎÊÌâת»¯ÎªÁ½²ãMDPÎÊÌ⣬£¬£¬±¾ÎÄÊ×´ÎÌá³öÁËMDPµÄCVaRÖ¸±êµÄÓÐÓÃÇó½âËã·¨£¬£¬£¬»®·Ö»ñµÃÁËÒ»ÖÖ¿É¿ìËÙÊÕÁ²ÖÁ¾Ö²¿×îÓŵÄÕ½ÂÔµü´úÐÍËã·¨ÒÔ¼°Ò»ÖÖ»ùÓÚѸËÙ¶ÈÆÊÎöµÄÈ«¾Ö×îÓÅËã·¨£¬£¬£¬Ìî²¹ÁËÏÖÓÐMDPÎÄÏ×¹ØÓÚCVaRµÄÓÐÓÃÇó½âËã·¨µÄ¿Õȱ¡£

ÂÛÎÄÁ´½Ó £º£º£ºhttps://doi.org/10.1111/poms.14077


¡¾ÍøÕ¾µØÍ¼¡¿