摘要
本文根据解决问题需要一定相关信息的基本原理提出了信息域概念,提出了一组关于信息域的基本假定,建立了信息域前后向关联矩阵及其有关结论,并根据模拟退火原理给出了具有创造性的联想算法CRA,讨论了其中的若干关键问题,如初始状态的选择,过程稳定性的检验,降温的策略,循环次数的确定,以及算法终止条件的设置等。该联想算法使得信息关联度低于阈值的靶仍然能以较小的概率被想起。理论与实验证明CRA正确地反映了联想思维的某些创造性,克服了已有模型的不足。
For solving any problem, the related information has to be provided. Hence, this paper firstly presents elementary hypotheses and establishes relevance matrix on information field, and then a creative reminding algorithm CRA, which based on the simulated annealing-based theory, is presented so as that some new cases whose information relevance degree less than the given threshold are still able to be reminded. Some key techniques such as selection of initial states, termination condition and so forth are also coped with. Theory analysis and experiment come to light that CR.A reflects more accurately the creativity of reminding thought ignored by the known models.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1999年第4期393-401,共9页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金
教育部博士点基金
广东省自然科学基金
