非线性约束优化的算法分析

针对非线性约束优化问题,运用了一种新的智能优化算法——社会认知优化算法。社会认知优化算法是一种基于社会认知理论的集群智能优化算法,它对目标函数的解析性质没有要求,适合于大规模约束问题处理的优点,使搜索不容易陷入局部最优。将该算法引入非线性约束问题,解决优化问题。通过实例和其他算法进行比较,对比数值实验结果表明,即使只有一个学习主体,该算法能够高效、稳定地得到解决方案,便于求解非线性约束优化问题。     

求解非线性方程组的社会认知算法

将非线性方程组的求解问题转化为函数优化问题,应用一种新的智能优化算法——社会认知算法求解此优化问题,实验结果表明了社会认知算法在求解非线性方程组时的可行性和有效性。     

求解一类不可微多目标优化问题的社会认知算法*

针对一类不可微多目标优化问题,给出了一个新的算法——极大熵社会认知算法。利用极大熵方法将带有约束的不可微多目标优化问题转化为无约束单目标优化问题,然后利用社会认知算法对其进行求解。该算法是基于社会认知理论,通过一系列的学习代理来模拟人类的社会性和智能性从而完成对目标的优化。利用两个测试算例对其进行测试并与其他算法进行比较,计算结果表明,该算法在求解的准确性和有效性方面均优于其他算法。     

求解一类随机优化问题的社会认知算法

提出了解随机优化问题的社会认知算法.该算法易于理解及程序易实现,克服了随机优化问题难以高效实现全局优化的缺点,为随机优化问题的求解提供了一种新的途径,并为社会认知算法的应用拓展了新的空间.     

求解非线性互补问题的熵函数认知优化算法

提出了一个求解非线性互补问题的熵函数社会认知优化算法。首先将非线性互补问题转化为非线性方程组来求解,然后利用熵函数法将非线性方程组求解转化为一个光滑的无约束优化问题,最后应用社会认知优化算法求解此优化问题。实验结果表明,该算法收敛速度快,稳定性好,是求解非线性互补问题的一种有效算法。     

QoS感知的云服务优化组合研究

为了更好地解决服务质量感知的云服务优化组合问题,首先对社会认知算法进行了改进,提出了面向离散型优化问题的模仿学习方法以及基于多次变异的观察学习方法。然后使用改进的社会认知算法对服务质量感知的云服务优化组合问题进行了求解。实验结果表明,改进的社会认知算法具有较强的搜索能力和较快的收敛速度,并且具有较强的推广性,可以用来求解其他离散型优化问题。     

基于文化社会认知算法的云服务优化组合研究

QoS感知的云服务优化组合是云计算领域亟需解决的问题。针对该问题,首先对社会认知算法进行了改进,然后将改进的社会认知算法纳入文化算法的框架之内,构造了新颖的文化社会认知算法,并采用该算法解决QoS感知的云服务优化组合问题。实验结果表明,文化社会认知算法在求解云服务优化组合问题时具有较强的搜索能力和较快的收敛速度;同时,该算法具有很强的推广性,可以用来求解其他类似的组合优化问题。     

一类非线性极大极小问题的极大熵社会认知算法

针对一类非线性极大极小问题目标函数非光滑的特点给求解带来的困难,利用社会认知算法并结合极大熵函数法给出了此类问题的一种新的有效算法。首先利用极大熵函数将原问题转化为一个光滑无约束优化问题,然后利用社会认知算法对其进行求解。该算法是基于社会认知理论,通过一系列的学习代理来模拟人类的社会性以及智能性从而完成对目标的优化。数值结果表明,该算法收敛快,数值稳定性好,是求解非线性极大极小问题的一种有效算法。     

强社会认知能力的粒子群优化算法

针对粒子群优化算法的“早熟”问题,提出了强社会认知能力粒子群优化算法,该算法通过学习概率和选择概率确定粒子跟踪的局部极值。算法中学习概率的自适应调整有效权衡了粒子的个体认知能力和社会认知能力。通过经典函数的测试结果表明,新算法的全局搜索能力有了显著提高,并且能够有效避免早熟问题。     

求解互补问题的极大熵社会认知算法

针对传统算法无法获得互补问题的多个最优解的困难,提出了求解互补问题的社会认知优化算法.通过利用NCP函数,将互补问题的求解转化为一个非光滑方程组问题,然后用凝聚函数对其进行光滑化,进而把互补问题的求解转化为无约束优化问题,利用社会认知算法对其进行求解.该算法是基于社会认知理论,通过一系列的学习代理来模拟人类的社会性以及智能性从而完成对目标的优化.该算法对目标函数的解析性质没有要求且容易实现,数值实验结果表明了该方法是有效的.     

Development of a Composite Knowledge Manipulation Tool: K-Expert

A goal of this study is to develop a Composite Knowledge Manipulation Tool (CKMT). Some of traditional medical activities are rely heavily on the oral transfer of knowledge, with the risk of losing important knowledge. Moreover, the activities differ according to the regions, traditions, experts’ experiences, etc. Therefore, it is necessary to develop an integrated and consistent knowledge manipulation tool. By using the tool, it will be possible to extract the tacit knowledge consistently, transform different types of knowledge into a composite knowledge base (KB), integrate disseminated and complex knowledge, and complement the lack of knowledge. For the reason above, I have developed the CKMT called as K-Expert and it has four advanced functionalities as follows. Firstly, it can extract/import logical rules from data mining (DM) with the minimum of effort. I expect that the function can complement the oral transfer of traditional knowledge. Secondly, it transforms the various types of logical rules into database (DB) tables after the syntax checking and/or transformation. In this situation, knowledge managers can refine, evaluate, and manage the huge-sized composite KB consistently with the support of the DB management systems (DBMS). Thirdly, it visualizes the transformed knowledge in the shape of decision tree (DT). With the function, the knowledge workers can evaluate the completeness of the KB and complement the lack of knowledge. Finally, it gives SQL-based backward chaining function to the knowledge users. It could reduce the inference time effectively since it is based on SQL query and searching not the sentence-by-sentence translation used in the traditional inference systems. The function will give the young researchers and their fellows in the field of knowledge management (KM) and expert systems (ES) more opportunities to follow up and validate their knowledge. Finally, I expect that the approach can present the advantages of mitigating knowledge loss and the burdens of knowledge transformation and complementation.     

基于Rough Set理论的一种属性值约简算法

属性值的约简是Rough Set理论的核心内容之一。它的口的就是在保持规则集的分类能力的条件下，删除多余属性值，进一步简化规则集。从而，得到最小的知识库。本文针对Rough Set理论中值约简这个重要问题进行了研究，提出了一种利用决策规则质量的属性值约简算法。该算法比现有的值约简算法更简化，并用实验证明了其有效性。     

Eliciting Implicit Assumptions of Mizar Proofs by Property Omission

When formalizing proofs with proof assistants, it often happens that background knowledge about mathematical concepts is employed without the formalizer explicitly requesting it. Such mechanisms are warranted in the context of discovery because they can make prover sessions more efficient (less time searching the library) and can compress proofs (the more knowledge that is implicitly available, the less needs to be explicitly said by the formalizer). In the context of justification, though, implicit knowledge may need to be made explicit. To study implicit knowledge in proof assistants, one must first characterize what implicit knowledge should be made explicit. Then, once a class of implicit background knowledge is identified, one needs to determine how to extract it from proofs. When a class of implicit knowledge is made explicit, we may then inquire to what extent the implicit knowledge is needed for any particular proof; it often happens that proofs can be successful even if some of the implicit knowledge is omitted. In this note we describe an experiment conducted on the Mizar Mathematical Library (MML) of formal mathematical proofs concerning a particular class of implicit background knowledge, namely, properties of functions and relations (e.g., commutativity, asymmetry, etc.). In our experiment we separate, for each theorem of the MML, the needed function and relation properties from the unneeded ones. Special attention is paid to those function and relation properties that are significant in discussions of foundations of mathematics.     

The conundrum of IT management

This paper presents a robust argument as to why it can be difficult for chief information officers (CIOs) to generate business value from investments that their organizations make in information technology (IT) with contemporary organizational structures, authority patterns, processes and mindsets. This argument is built on the subtle premise that organizations should not seek to merely manage IT but to manage the delivery of business value through IT. It takes the view that this latter quest is knowledge-based and that the knowledge resources to successfully deliver this value are distributed throughout the organization. Crucially, this knowledge is not located solely within the IT function, presenting a challenge for the CIO for its integration and coordination. With the CIO having little or no jurisdiction over all required knowledge, its deployment will therefore be fragmented. The conundrum of IT management is how to generate value through IT without having access and authority over necessary resources. Research and practitioner implications of this analysis are highlighted.     

基于密码库函数的程序加解密过程分析

针对常见密码算法种类多及实现方式不同,采用现有特征扫描和动态调试的方法分析程序中的加解密过程非常困难的问题。提出一种基于库函数原型分析和库函数调用链构造的加解密过程分析方法,库函数原型分析是分析常见密码库函数所包含的密码算法知识和库框架知识,并记录形成知识库,库函数调用链是根据密码库函数调用时参数值的相等关系构建的库函数调用的先后关系链,最后根据知识库在链上提取展示密码库及密码算法相关知识。该方法对运用到常见库的程序中的算法的识别精确度达到近100%,能详细分析算法调用时的数据、密钥、模式,并有助于对多个算法的协同处理关系作分析。该方法有助于辅助分析木马、蠕虫之类恶意程序,也可用于检测程序对库密码算法的运用是否正确。     

Computing with the social fabric: The evolution of social intelligence within a cultural framework

Cultural algorithms employ a basic set of knowledge sources, each related to knowledge observed in various animal species. These knowledge sources are then combined to direct the decisions of the individual agents in solving optimization problems using an influence function inspired by the marginal value theorem from population biology. We briefly describe an implementation of this approach, the cultural algorithms toolkit (CAT) in the Repast agent-based simulation environment. Next we introduce the notion of "social fabric" which provides a framework in which the knowledge sources can access the social networks to which individuals can belong. A computational version of the social fabric idea is then implemented as an extension of the influence function in the CAT system. We then apply the enhanced system to a problem in engineering design, the "pressure vessel problem". For this problem, we show that the enhanced system with the social fabric outperforms the CAT system without it. We demonstrate also that the frequency with which the knowledge sources are able to access the network can affect the problem solving process.     

The search for knowledge,contexts, and Case-Based Reasoning

A major goal of this paper is to compare Case-Based Reasoning with other methods searching for knowledge. We consider knowledge as a resource that can be traded. It has no value in itself; the value is measured by the usefulness of applying it in some process. Such a process has info-needs that have to be satisfied. The concept to measure this is the economical term utility. In general, utility depends on the user and its context, i.e., it is subjective. Here, we introduce levels of contexts from general to individual. We illustrate that Case-Based Reasoning on the lower, i.e., more personal levels CBR is quite useful, in particular in comparison with traditional informational retrieval methods.     

一种适用于IDS的多次模糊迭代特征选择算法

本文针对入侵检测系统（IDS）被检测数据的特点，对适用于IDS的特征选择算法进行了研究，提出了一种基于分类的多次模糊迭代特征选择算法。该算法包括在属性空间中搜索特征子集、评估每个候选特征子集和分类这3个步骤，设计了与之相应的搜索算法和评估函数；算法通过多次迭代去除特征值集的冗余特征，得到精确度较高的特征值集；使用模糊逻辑得到与精确度要求相应的取值范围；由于单纯对数据进行操作，能比依赖于领域知识的算法更客观地分析数据。文内还对所提出的算法做了测试实验；并将实验结果与用可视化工具产生的特征可视化结果进行了比较。结果表明：该算法在IDS数据集上可取得良好的特征选择效果。     

Least squares approximate policy iteration for learning bid prices in choice-based revenue management

We consider the revenue management problem of capacity control under customer choice behavior. An exact solution of the underlying stochastic dynamic program is difficult because of the multi-dimensional state space and, thus, approximate dynamic programming (ADP) techniques are widely used. The key idea of ADP is to encode the multi-dimensional state space by a small number of basis functions, often leading to a parametric approximation of the dynamic program’s value function. In general, two classes of ADP techniques for learning value function approximations exist: mathematical programming and simulation. So far, the literature on capacity control largely focuses on the first class.In this paper, we develop a least squares approximate policy iteration (API) approach which belongs to the second class. Thereby, we suggest value function approximations that are linear in the parameters, and we estimate the parameters via linear least squares regression. Exploiting both exact and heuristic knowledge from the value function, we enforce structural constraints on the parameters to facilitate learning a good policy. We perform an extensive simulation study to investigate the performance of our approach. The results show that it is able to obtain competitive revenues compared to and often outperforms state-of-the-art capacity control methods in reasonable computational time. Depending on the scarcity of capacity and the point in time, revenue improvements of around 1% or more can be observed. Furthermore, the proposed approach contributes to simulation-based ADP, bringing forth research on numerically estimating piecewise linear value function approximations and their application in revenue management environments.     

A novel model by evolving partially connected neural network for stock price trend forecasting

This paper proposes a novel model by evolving partially connected neural networks (EPCNNs) to predict the stock price trend using technical indicators as inputs. The proposed architecture has provided some new features different from the features of artificial neural networks: (1) connection between neurons is random; (2) there can be more than one hidden layer; (3) evolutionary algorithm is employed to improve the learning algorithm and training weights. In order to improve the expressive ability of neural networks, EPCNN utilizes random connection between neurons and more hidden layers to learn the knowledge stored within the historic time series data. The genetically evolved weights mitigate the well-known limitations of gradient descent algorithm. In addition, the activation function is defined using sin(x) function instead of sigmoid function. Three experiments were conducted which are explained as follows. In the first experiment, we compared the predicted value of the trained EPCNN model with the actual value to evaluate the prediction accuracy of the model. Second experiment studied the over fitting problem which occurred in neural network training by taking different number of neurons and layers. The third experiment compared the performance of the proposed EPCNN model with other models like BPN, TSK fuzzy system, multiple regression analysis and showed that EPCNN can provide a very accurate prediction of the stock price index for most of the data. Therefore, it is a very promising tool in forecasting of the financial time series data.