【摘要】 运用函数连接型神经网络研究 3 8种取代芳烃化合物的分子结构—毒性的定量构效关系 (QSAR) ,计算结果与MLR计算结果比较 ,前者比后者好。与实验结果的线性拟合相关系数 ,FLN所得结果为 0 9561～ 0 9996,MLR所得结果为 0 869～ 0 965。更多还原

【Abstract】 Systematic studies were made on the applications of the Functional\|Link Net (FLN), a novel single\|layer neural network, without hidden neurons, to the relationship between the Na\|+\|K\++ adenosine triphosphatase activities and the chemical structures of substituted aromatic compounds. The nonlinearity of original input patterns is enhanced, and the generalized δ\|learning rule is used in FLN. By comparing the FLN with multiple linear regression models (MLR), it was found that the former behaves better than the latter. The correlation coefficients, 0 9561～0 9996, of the results obtained by FLN is better than those of the results obtained by MLR, 0 869～0 965.更多还原