基于二进制神经网络集成的潜在通路分析

针对神经网络在潜在通路分析应用中的缺陷,提出了二进制神经网络集成(BNNE)算法。该算法结合了二进制神经网络(binary neural network,BNN)和神经网络集成(neural network ensemble,NNE),NNE的个体成员是BNN,集成算法采用GASEN算法,其输入为电路开关状态,输出为预测功能。通过比较特定开关状态下的预测功能和设计功能之间的差异,判断是否存在潜在通路。该算法综合了BNN、NNE的优点,可以有效提高神经网络的泛化能力和潜在通路分析的可靠性,仿真试验验证了算法的有效性。

Binary neural network ensemble for sneak circuit analysis

To overcome the shortcomings of the sneak circuit analysis(SCA) based on BPNN,a binary neural network ensemble(BNNE) algorithm for the sneak circuit analysis(SCA) is proposed.Individuals in an ensemble are binary neural networks(BNNs),in which inputs and outputs are binary representation.BNNE solves the problem of selection of threshold and have the better generalization performance than individual BPNN.The inputs of BNNE are the states of switches in circuit,and the outputs are predicted functions.The trained BNNE can predict the all possible functions of the circuit.The sneak circuits are discovered by comparison between the predicted functions and desired functions.BNNE combines BNN and NNE to improve the performance of SCA.The paper provides experimental evidence that supports the algorithm.