基于鱼群算法的极限学习机影像分类方法优化

在传统极限学习机(ELM)研究的基础上,考虑到传统ELM参数的不确定会导致整体分类精度下降,利用仿生鱼群算法(AF)对ELM的小波核参数和正则化参数进行寻优,并构造参数优化后的小波ELM影像分类模型(AF-ELM)。通过实验比较了该算法与人工神经网路(ANN)、支持向量机(SVM)、极限学习机(ELM)等标准分类器在遥感影像分类上的精度与速度差异,并且与ELM多项式核、RBF核分类算法进行比较分析,验证了AF-ELM在分类速度和精度上的优越性。实验结果表明,AF-ELM分类方法分类速度较快,精度较高,均优于其他分类方法。能较好地应用于遥感影像上各类地物要素的自动提取。     

Age, differential growth and mortality rates in unexploited populations of Florida gar, an apex predator in the Florida Everglades

Abstract  Florida gar, Lepisosteus platyrhincus DeKay, were sampled in two canal systems in south Florida during 2000–2001 to estimate age, growth and mortality as part of the Everglades ecosystem-restoration effort. Tamiami (C-4) and L-31W canal systems had direct connections to natural wetlands of the Everglades and harboured large Florida gar populations. Of 476 fish aged, maximum ages were 19 and 10 years for females and males, respectively. Maximum sizes were also larger for females compared with males (817 vs 602 mm total length). Overall, female Florida gar from both Tamiami and L-31W were larger at age than males from L-31W that, in turn, were larger at any given age than males from Tamiami. Females also had lower rates of annual mortality ( Z  = 0.21) than males from L-31W ( Z  = 0.31) or males from Tamiami ( Z  = 0.54). As a large and long-lived apex predator in the Everglades, Florida gar may structure lower trophic levels. Regional- and sex-specific population parameters for Florida gar will contribute to the simulation models designed to evaluate Everglades restoration alternatives.