| 摘要: |
| 提出了一个新的改进型指数双向联想记忆模型(improved eBAM,简称IeBAM).通过定义有界且随状态改变而下降的能量函数,证明了IeBAM在状态的同、异步更新方式下的稳定性,一方面排除了Wang的修正指数BAM(modified eBAM,简称MeBAM)和Jeng的eBAM(exponential BAM)的稳定性证明中所作的不合理假设;另一方面,放宽了对BAM(bidirectional associative memory)的连续性假设的要求,并避免了补码问题.理论分析和计算机模拟结果表明, |
| 关键词: 神经网络,（指数）联想记忆,稳定性,双向性,性能估计. |
| DOI： |
| 分类号: |
| 基金项目:本文研究得到国家自然科学基金资助. |
|
| Improved Exponential Bidirectional Associative Memory and Its Performance Evaluation |
|
CHEN Song-can,GAO Hang
|
| Abstract: |
| In this paper, a new improved exponential bidirectional associative memory (IeBAM) model is proposed. Its stability in synchronous and asynchronous updating modes of the states is proven by defining an energy function which is bounded and decreases as the states change. On one hand, IeBAM eliminates the unreasonable hypotheses in the stability proofs of both Wang's modified exponential BAM (MeBAM) and Jeng's exponential BAM (eBAM). On the other hand, it relaxes the continuity assumption of the BAM (bidirectional associative memory) and avoids the complement encoding problem. The theoretical analysis and computer simulations indicate that the IeBAM has higher storage capacity and better error-correcting capability than the MeBAM and the eBAM. |
| Key words: Neural networks, (exponential) associative memories, stability, bidirectionality, performance evaluation. |
