| 利用人工鱼群算法优化高斯过程模型及应用分析 |
| |
| 引用本文： | 邱小梦,周世健,王奉伟,欧阳亮酉. 利用人工鱼群算法优化高斯过程模型及应用分析[J]. 测绘通报, 2018, 0(1): 157-160,164. DOI: 10.13474/j.cnki.11-2246.2018.0031 |
| |
| 作者姓名： | 邱小梦 周世健 王奉伟 欧阳亮酉 |
| |
| 作者单位： | 1. 东华理工大学测绘工程学院, 江西 南昌 330013;2. 流域生态与地理环境监测国家测绘地理信息局重点实验室, 江西 南昌 330013;3. 南昌航空大学, 江西 南昌 330063;4. 同济大学测绘与地理信息学院, 上海 200092 |
| |
| 基金项目： | 国家自然科学基金，江西省研究生创新基金 |
| |
| 摘 要： | 基于高斯过程利用传统共轭梯度法搜索超参数，存在对初始值依赖性强、获得局部最优解的不足。本文采用人工鱼群算法对超参数进行智能寻优，建立了基于人工鱼群算法的高斯过程模型对变形体形变进行预测分析。通过隧道和基坑两个工程实例计算比对分析，NN、SE和RQ 3种核函数中NN核函数的预测效果最好，平均相对误差分别为0.69%和1.06%。结果表明超参数优化模型的预测精度得到了较大的提高，改善了高斯过程算法本身存在的超参数求解方面的不足，效果明显。
|
| 关 键 词： | 人工鱼群算法 高斯过程 滚动预测法 变形监测 |
| 收稿时间： | 2017-04-19 |
| 修稿时间： | 2017-06-05 |
Optimization of Gaussian Process Model Based on Artificial Fish School Algorithm and Its Application Analysis |
| |
| QIU Xiaomeng,ZHOU Shijian,WANG Fengwei,OUYANG Liangyou. Optimization of Gaussian Process Model Based on Artificial Fish School Algorithm and Its Application Analysis[J]. Bulletin of Surveying and Mapping, 2018, 0(1): 157-160,164. DOI: 10.13474/j.cnki.11-2246.2018.0031 |
| |
| Authors: | QIU Xiaomeng ZHOU Shijian WANG Fengwei OUYANG Liangyou |
| |
| Affiliation: | 1. Faculty of Geomatics, East China University of Technology, Nanchang 330013, China;2. Key Laboratory of Watershed Ecology and Geographical Environment Monitoring, NASMG, Nanchang 330013, China;3. Nanchang Hangkong University, Nanchang 330063, China;4. College of Surveying and Geoinformatics, Tongji University, Shanghai 200092, China |
| |
| Abstract: | Based on the Gaussian process,the traditional conjugate gradient method is used to search the hyper-parameters, which has the disadvantages of strong dependence on the initial value and obtain local optimal solution.Artificial fish swarm algorithm is used to optimize the hyper-parameters intelligently, and the Gaussian process model based on the artificial fish swarm algorithm is used to predict and analysize the deformation of deformable body.Through the tunnel and foundation pit examples to compare the calculated results, NN kernel function is the best predictor of NN, SE and RQ, the average relative error is 0.69％and 1.06％.The results show that the prediction accuracy of the hyper-parameters optimization model are greatly improved, and the hyper-parameters solving of the Gaussian process algorithm themselves are improved and the effect is obvious. |
| |
| Keywords: | artificial fish swarm algorithm Gaussian process rolling prediction method deformation monitoring |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《测绘通报》浏览原始摘要信息 |
|
点击此处可从《测绘通报》下载全文 |
|
