Efficiency Bound of Learning with Coarse Graining  

作　　者：李明昊 夏世豪 王有林 律明龙 陈金灿 苏山河 Minghao Li;Shihao Xia;Youlin Wang;Minglong Lv;Jincan Chen;Shanhe Su(Department of Physics,Xiamen University,Xiamen 361005,China)

机构地区：[1]Department of Physics,Xiamen University,Xiamen 361005,China

出　　处：《Chinese Physics Letters》2023年第11期31-36,共6页中国物理快报（英文版）

基　　金：supported by the National Natural Science Foundation of China(Grant No.12075197);the Fundamental Research Fund for the Central Universities(Grant No.20720210024);the Natural Science Foundation of Fujian Province(Grant No.2023J01006)。

摘　　要：A thermodynamic formalism describing the efficiency of information learning is proposed,which is applicable to stochastic thermodynamic systems with multiple internal degrees of freedom.The learning rate,entropy production rate and entropy flow from the system to the environment under coarse-grained dynamics are derived.The Cauchy–Schwarz inequality is applied to demonstrate the lower bound on the entropy production rate of an internal state.The inequality of the entropy production rate is tighter than the Clausius inequality,leading to a derivative of the upper bound on the efficiency of learning.The results are verified in cellular networks with information processes.

关 键 词：STATE INEQUALITY ENTROPY 

分 类 号：O414.1[理学—理论物理]