GOS+MU:一种查询对象选择新方法

在分析单一MU(Most Uncertainty)采样缺陷的基础上,提出一种"全局最优搜寻"方法 GOS(Global Optimal Search),并结合MU共同完成查询选择。GOS+MU方法中,GOS着眼全局寻找目标,在应用环境能提供的训练样本数量有限、分类器受训不充分时,该方法选择的对象学习价值高,能快速推进分类器学习进程;MU则能够在GOS采样失效情形下,利用分类器当前训练成果,选择查询不确定性最强的样本补充训练集。通过对网络商品的用户评论进行分类仿真,并比较其他采样学习方法的效果,证明了GOS+MU方法在压缩学习成本、提高训练效率方面的有效性。     

Prediction of methane adsorption content in continental coal-bearing shale reservoir using SLD model

Shale gas, as an important unconventional resource, has drawn global attention. It is mainly composed of adsorption gas and free gas. Adsorption gas content could play an important guiding role on both the selection of favorable perspective area and the exploration and exploitation of shale gas resources. In order to accurately measure adsorption gas content, a new approach was established to predict the adsorption isotherm of methane on shale. Based on the simplified local-density (SLD) method, both the adsorption isotherms of illite, illite/smectite mixed-layer, cholorite and type III kerogen and the total shale rock could be well fitted. The fitting results show good coincidences with the true experimental test data, which proves the method is reasonable and dependable and the prediction results are effective and credible. In addition, the good simulation results show that the SLD parameters can reflect the pore structure characteristics and corresponding adsorption characteristics of the shale samples, which can be used for the quantitative characterization of shale pore system.