Abstract: Aiming at the flaw that the current cognitive computations aren’t able to simulate cognition accurately, the cognitive computation based on cognitive unit is proposed. All the cognitive data originate from cognitive unit. The cognitive computation based on cognitive unit is composed of producing cognitive data, storing cognitive data and improving cognitive data. Experimental results demonstrate that the cognitive computation based on cognitive unit accords closely with cognitive rules of human beings, and increasing the memory volume improves all the cognitive computation data. The achievement rate of work and the rate of losing cognition reach optimal values while the memory volume is large enough, but the rate of self-examination cognition and the rate of deepening cognition only reach saturation values. So the cognitive computation based on cognitive unit is able to simulate cognition accurately.

Key words: cognitive unit, cognition, work, cognitive computation data

摘要： 为改正已有认知计算不能准确模拟人类认知的缺陷，提出了基于认知元的认知计算。它的认知数据来源于认知元，认知计算由生成认知数据、存储认知数据、优化认知数据组成。研究发现，基于认知元的认知计算符合人类的认知规律；增加存储体容量能够改善认知计算的所有性能指标，当存储体的容量增加到足够大时，成功率及遗忘率可以达到最理想的值，但深化率和反思率却只能达到饱和值。因此，基于认知元的认知计算能够准确模拟人类的认知。

关键词: 认知元, 认识, 任务, 认知计算指标