极限学习机矿产资源评价研究

项目名称： 极限学习机矿产资源评价研究

项目编号： No.41272360

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 天文学、地球科学

项目作者： 陈永良

作者单位： 吉林大学

项目金额： 85万元

中文摘要： 极限学习机是一种学习速度快、操作容易、人工干预少、泛化性能强的新型计算智能模型，已成为大规模计算和人工智能的有效理论工具。将该模型引入矿产资源评价研究领域，有望解决多元非线性统计模型在矿产资源评价应用中面临的计算复杂度过高和存储空间过大等诸多技术瓶颈问题。鉴于此，项目组拟概括和总结极限学习机理论与应用研究的最新成果，借鉴极限学习机和极限学习机集成模型在解决大规模并行回归和分布式集成分类等问题方面的应用研究经验，研制基于极限学习机、并行式极限学习机集成模型、在线顺序极限学习机和分布式在线顺序极限学习机集成模型的一系列多元非线性矿产资源评价模型；以实验区地物化遥多源地学观测数据为基本数据源，应用新模型预测矿产资源远景靶区，创建极限学习机矿产资源评价应用研究的一个典型范例。研究工作能够为当代矿产资源评价提供新技术与新方法，对极限学习机理论的发展具有积极的促进作用，理论与实际意义重大。

中文关键词： 多源地学观测数据；地质统计单元；致矿地质异常；极限学习机；矿产资源评价

英文摘要： Extreme learning machine is a new computational intelligence model with fast learning speed, easy implementation, few human intervene, and high generalization performance. It becomes an effective theoretical tool for large scale computation and artificial intelligence. By introducing this model into mineral resource assessment, it is expected to solve a number of technical bottleneck problems, such as too high computational complexity, too large memory space, and etc., faced in the applications of nonlinear multivariate statistical models in mineral resource assessment. For this reason, we plan to generalize and summarize the latest research achievements of extreme learning machine on theories and applications. By learning from the experiences of extreme learning machine and ensemble of extreme learning machines used in large-scale parallelized regression and distributed ensemble classification and etc., we plan to innovate a set of nonlinear multivariate statistical models for mineral resource assessment based on extreme learning machine, parallelized ensemble of extreme learning machines, online sequential extreme learning machine, and distributed ensemble of online sequential extreme learning machines. Using the multisource geo-observation data including geological, geochemical, geophysical, and remote sensin

英文关键词： Multisource geo-obsevration data；Geostatistical cell；mineralization associated geo-anomaly；Extreme learning machine；Mineral resource assessment