偏最小二乘回归的贝叶斯正则化神经网络集成模型在证券分析预测中的应用

神经网络集成技术能有效地提高神经网络的预测精度和泛化能力,已经成为机器学习和神经计算领域的一个研究热点.利用Bagging技术和不同的神经网络算法生成集成个体,并用偏最小二乘回归方法从中提取集成因子,再利用贝叶斯正则化神经网络对其集成,以此建立上证指数预测模型.通过上证指数开、收盘价进行实例分析,计算结果表明该方法预测精度高、稳定性好.     

基于BP神经网络的企业未来获利能力智能综合评价

分析了相关分析——多指标综合评价法在确定企业未来获利能力方面的优点和不足 ;并在其基础上提出了基于 BP神经网络的多指标综合评价法 ;仿真试验证明了基于 BP神经网络的多指标综合评价法的有效性     

人工神经网络在SARS疫情分析与预测中的应用

讨论人工神经网络在 SARS疫情分析与预测中的应用 .采用三层结构的反向传播网络 ( Backpropagation network,简称 BP网络 ) ,对 SARS在中国的传播与流行趋势及控制策略建立了网络模型 .并利用实际数据拟合参数 ,针对北京、山西的疫情进行了计算仿真 .结果表明 ,该网络模型算法收敛速度较快 ,预测精度很高     

Ring‐opening grafting polymerization of cyclic monomers onto human hair

Natural human hair was successfully modified by the graft polymerization of trimethylene carbonate, β‐propiolactone, ε‐caprolactone, glycidol, ε‐caprolactam, and 5,5‐dimethyl‐1,3‐dioxane‐2‐thione. In contrast, we could not modify natural human hair by the graft polymerization of oxetane under similar conditions. The model reaction suggested that the main initiating species in these polymerizations were the amino, thiol, and hydroxyl groups in hair, which could induce ring‐opening polymerization. Among the tested monomers, β‐propiolactone was most effective for hair modification with its graft polymer, whose concentration was as high as 0.5 g/g of hair though polymerization under mild conditions. The effects of the hair pretreatment and polymerization temperature on the weight ratio of the grafted polymers were also investigated. Hair modified by grafted polymers was characterized with scanning electron microscopy and Fourier transform infrared measurements. © 2007 Wiley Periodicals, Inc. J Polym Sci Part A: Polym Chem 45: 736–744, 2007     

电力市场中合同电量与竞争电量交易比例的研究

在单边开放的区域电力市场中,合理的合同电量与竞争电量交易比例是保证电力市场有效运行的一个重要环节。竞争电量所占的比例将主要取决于当前发电公司的市场行为。首先使用BP神经网络对电力需求弹性系数进行了预测,然后以长期电力市场均衡为目标函数,考虑贵州电网发电机组的可用容量与负荷预测的误差,以及贵州输电线路的可靠性诸因素,推导出合同电量与竞争电量交易比例,经过与南方区域电力市场目前运营规则规定的交易比例比较,该比例是合理的,可以规避电力市场价格波动等带来的风险。     

A novel approach based on recurrent neural networks applied to nonlinear systems optimization

This paper presents an efficient approach based on recurrent neural network for solving nonlinear optimization. More specifically, a modified Hopfield network is developed and its internal parameters are computed using the valid subspace technique. These parameters guarantee the convergence of the network to the equilibrium points that represent an optimal feasible solution. The main advantage of the developed network is that it treats optimization and constraint terms in different stages with no interference with each other. Moreover, the proposed approach does not require specification of penalty and weighting parameters for its initialization. A study of the modified Hopfield model is also developed to analyze its stability and convergence. Simulation results are provided to demonstrate the performance of the proposed neural network.     

On Numerical Solution of Shape Inverse Problems

The new method is proposed for the numerical solution of a class of shape inverse problems. The size and the location of a small opening in the domain of integration of an elliptic equation is identified on the basis of an observation. The observation includes the finite number of shape functionals. The approximation of the shape functionals by using the so-called topological derivatives is used to perform the learning process of an artificial neural network. The results of computations for 2D examples show, that the method allows to determine an approximation of the global solution to the inverse problem, sufficiently closed to the exact solution. The proposed method can be extended to the problems with an opening of general shape and to the identification problems of small inclusions. However, the mathematical theory of the proposed approach still requires futher research. In particular, the proof of global convergence of the method is an open problem.     

手分为多节段的人体热调节模型研究

将人体热调节柱状模型和手的多节段模型结合起来,建立了更完整的人体热调节模型。利用有限元方法对人体热调节数学模型进行了数值求解,并设计试验证明了模型的正确性。结果表明: (1)当取合适的血流量时,实验值和计算值的变化趋势一致,且手部温度稳定时,血流量都在确定的范围内;(2)血液流量是手部温度场变化的主要因素,人体组织和动脉入口温度对手部温度影响较小。     

基于模糊神经网络的自适应PID控制器

提出了一种应用于自适应PID控制器的神经网络与模糊控制相结合的算法,该算法可以有效地解决普通PID控制器依赖于对象的数学模型的缺点,可实现控制系统的在线自适应调整,可满足实时控制的要求。仿真结果表明,基于模糊神经网络整定的PID控制器具有较好的自学习和自适应性,具有较快的响应速度。     

Modeling of environmentally significant interfaces: Two case studies

When some parameters cannot be easily measured experimentally, mathematical models can often be used to deconvolute or interpret data collected on complex systems, such as those characteristic of many environmental problems. These models can help quantify the contributions of various physical or chemical phenomena that contribute to the overall behavior, thereby enabling the scientist to control and manipulate these phenomena, and thus to optimize the performance of the material or device. In the first case study presented here, a model is used to test the hypothesis that oxygen interactions with hydrogen on the catalyst particles of solid oxide fuel cell anodes can sometimes occur a finite distance away from the triple phase boundary (TPB), so that such reactions are not restricted to the TPB as normally assumed. The model may help explain a discrepancy between the observed structure of SOFCs and their performance. The second case study develops a simple physical model that allows engineers to design and control the sizes and shapes of mesopores in silica thin films. Such pore design can be useful for enhancing the selectivity and reactivity of environmental sensors and catalysts. This paper demonstrates the mutually beneficial interactions between experiment and modeling in the solution of a wide range of problems.