基于条件生成式对抗网络的油藏单井产量预测模型

为克服机器学习方法在油藏单井产量预测中的过拟合问题,提高油田生产中的产量预测精度,提出一种基于条件生成式对抗网络(CGAN)的油藏单井产量预测模型。该模型使用长短期记忆、全连接等基础神经网络,构建生成和判别网络模型。生成网络模型以产量影响因素为条件输入,生成预测产量数据,利用对数损失函数评价预测数据与真实数据之间的偏差,通过条件生成式对抗网络的博弈训练,并结合贝叶斯超参数优化算法,优化模型结构,综合提高模型的泛化能力。基于Eclipse数值模拟软件建立同一井网条件下不同地质和生产条件下的油藏单井产量数据库,以地质与生产条件等产量影响因素作为模型的条件输入,进行油藏单井产量预测。结果表明：与全连接神经网络(FCNN)、随机森林(RF)以及长短期记忆神经网络(LSTM)模型的预测结果相比,CGAN模型在测试集上的平均绝对百分比误差分别提升了2.59%、 0.81%以及1.72%,并且过拟合比最小(1.027)。说明CGAN降低了机器学习产量预测模型的过拟合程度,提高了模型的泛化能力与预测精度,验证了所提算法的优越性,对指导油田高效开发和保障我国能源战略安全具有重要意义。     

Four-component superlattice empirical pseudopotential method for InAs/GaSb superlattices

For the design of InAs/GaSb superlattice (SL) heterojunction infrared photodetectors with very low dark current we have extended the standard two-component superlattice empirical pseudopotential method (SEPM) and implemented a four-component model including interface layers. For both models, the calculated bandgap values for a set of SL samples are compared to bandgaps determined by photoluminescence measurements. While the bandgap resulting from the two-component model agrees well with experimental data for SL structures with individual layer thicknesses of 7 monolayers and more, we show that for SLs with thinner GaSb layers the four-component SEPM model is accurate, when the As-content in the interface and barrier layers is included in the model.     

指数分布场合下步进应力加速寿命试验的Bayes分析

当寿命分布为指数分布时，本文给出了步进应力加速寿命试验的一种新的Ｂａｙｅｓ估计，为计算上的方便，本文还给出一种近似方法，数值例子表明，此方法简便可行。     

Characterization of limits of Bayes procedures

For the problem of estimating the natural parameter of a p-dimensional exponential family, a characterization of regular limits of Bayes procedures is obtained which generalizes results of Sacks [14], Brown [3], and Berger and Srinivasan [1]. The form is deduced under regularity conditions for the loss function which are more general than squared error. As a corollary it is then stated that the class of procedures with this form is a complete class. The parameter space may be open, and when it is closed, the limits of Bayes procedures are generalized Bayes.     

About membership functions estimation

A logical approach to fuzzy sets method originated by Giles is developed. The infinitely many-valued logic t_ω is taken as basic. We accept, it is correct to use the strong conjunction by the logical analysis of the summation of fuzzy items. Under some broad conditions it is proved. that the sum of many fuzzy variables is a variable whose membership function is approximately equal to $\text{?(x) =}\text{max}\text{{1 ?}\text{1}\text{2}\text{c(x ? α)}{}^2\text{, 0}}$, where a and c are some constant parameters. A method of estimation of the unknown parameters is developed in a general case. The proposed fuzzy method coincides with the method of maximum likelihood if used in problems of classical mathematical statistics.     

Consistent Lyapunov exponent Estimation for one-dimensional dynamical systems

Summary  The author proves the consistency of a nearest neighbor estimator of the Lyapunov exponent for a general class of one-dimensional ergodic dynamical systems. The author shows that this estimator has good practical properties on a set of simulations.     

Extracting deep information from limited observations on an evolved social network

We provide empirical evidence that in a social network which evolves over time, it is possible to extract deep information about the system from limited observations. In this paper, we consider a simple piece of readily available evidence on access to financial services by individuals in the UK. Detailed statistical analysis has shown that the decisions of agents on whether or not to have a basic financial account such as a bank account is heavily influenced by other individuals on their social network. We consider a small amount of straightforward and readily accessible information. We deduce from this, using an agent-based model, the type of social network across which information and influence on behaviour flows between agents in this context. Specifically, we show that information appears to flow across a small world network.     

采用子带二值加权累积的经验格林函数提取

提出一种采用子带二值加权累积的海洋环境噪声互相关函数(NCF)提取经验格林函数(EGF)方法.首先将每一快拍NCF在频域划分为多个子带,每个子带内根据"累积后提取的EGF信噪比增加"的准则,确定各快拍NCF的加权系数为0或1,将各子带的加权累积结果谱白化后在频域拼接,再反傅里叶变换得到时域EGF.子带二值加权累积方法实...     

Bayes empirical Bayes estimation of a poisson mean

In the empirical Bayes (EB) decision problem consisting of squared error estimation of a Poisson mean, a prior distribution λ is placed on the gamma family of prior distributions to produce Bayes EB estimators which are admissible. A subclass of such estimators is shown to be asymptotically optimal (a.o.). The results of a Monte Carlo study are presented to demonstrate the favorable a.o. property of the Bayes EB estimators in comparison with other competitors.     

K/n(G)系统可靠性评定的多源验前信息融合方法

在多源验前信息的情况下,以k/n(G)系统为例,讨论了如何对系统的可靠性指标进行估计的问题.在多个验前信息源给出了系统可靠性指标点估计的情况下,利用多层Bayes方法及经验Bayes方法对这些数据进行融合,并给出系统可靠性指标的Bayes点估计,同时对系统可靠度的置信区间也作了讨论.仿真算例表明这种处理方法是合理有效的.