Learning and imprinting in stationary and non-stationary environment

摘要: The performance of the extended Bush-Mosteller learning and imprinting scheme developed previously is studied for stationary non-stationary stochastic environments. As a criterion average missing information level chosen. For environment approximate time course latter derived discussed, an exact symmetry in schemes proved, biological advantage processes, with respect to energy consumption, pointed out. proper shown be superior as adaptability novel environmental properties decreases exponentially time. optimal memory range system calculated function span during which changes significantly mean amplitude these occur.