森林自疏规律BPMSM模型与张氏模型的比较

COMPARISON FOR BP-MSM MODEL AND ZHANG’S MODEL IN STUDYING FOREST SELF-THINNING

森林自然稀疏机制一般是非线性的、动态的 ,人工神经网络具有逼近任意非线性映射的特性 ,这从理论上保证了其应用于森林自疏研究的可行性。本文在提出的基于改进单纯形法的神经网络模型 (BP MSM混合算法 )的基本原理和算法的基础上 ,结合山杨林、云南松林和杉木林自疏实例进一步分析了BP MSM混合算法与张氏模型在研究森林自疏规律上的效果优劣。森林自疏实例应用结果表明 ,当建立BP MSM混合算法的 3层 1∶5∶1网络结构模型时 ,其模拟效果明显优于张氏模型 ,残差平方和仅为张氏模型的 3 89%～ 2 7 16 % ,说明BP MSM混合算法应用于森林自疏规律研究是理想的 ,从而丰富了森林自然稀疏规律研究方法。

The models of forest self thinning are generally nonlinear and dynamic. The artificial neural network has the characteristic of expressing arbitrary nonlinear mapping, which provides theoritic feasibility for modeling forest self thinning law. Based on the principle and algorithms of the neural network model based modified simplex method (BP MSM mixed algorithms), this paper analyzed the effect of BP MSM mixed algorithms and Zhang's model for modeling forest self thinning further. The comparisons in forest self thinning of \%Populus tremula\% var. \%davidiana\% forest, \%Pinus yunnanensis\% forest and \%Cunninghamia lanceolata\% forest illustrated that the simulated effect of BP MSM mixed algorithms were superior to Zhang's model significently when establishing network structure of 1∶5∶1 by three layers. The results of forest self thinning examples showed the surplus square of BP MSM mixed algorithms were only 3.89%～27.16% of Zhang's model, which were satisfactory and its precision were higher. This study will enrich the simulating method of forest self thinning, but the network structure of BP MSM mixed algorithms is important by choosing the numbers of concealing layer and neural points.