Online support vector regression for reinforcement learning

The goal in reinforcement learning is to learn the value of state-action pair in order to maximize the total reward. For continuous states and actions in the real world, the representation of value functions is critical. Furthermore, the samples in value functions are sequentially obtained. Therefore, an online sup-port vector regression （OSVR） is set up, which is a function approximator to estimate value functions in reinforcement learning. OSVR updates the regression function by analyzing the possible variation of sup-port vector sets after new samples are inserted to the training set. To evaluate the OSVR learning ability, it is applied to the mountain-car task. The simulation results indicate that the OSVR has a preferable con- vergence speed and can solve continuous problems that are infeasible using lookup table.     

Agents and Lattice-valued Logic

In fuzzy set theory, instead of the underlying membership set being a two -valued set it is a multi-valued set that generally has the structure of a lattice L with a minimal element O and the maximal element I. Furthermore if ∧, ∨, → and ┐ are defined in the set L, then we can use these operations to define, as in the ordinary set theory, operations on fuzzy subsets. In this paper we give a model of the Lattice-Valued Logic with set of agents.Any agents know the logic value of a sentence p. The logic value is compatible with all of the accessible conceptual models or worlds of p inside the agent. Agent can be rational or irrational in the use of the logic operation.Every agent of n agents can have the same set of conceptual models for p and know the same logic for p in this case the agents form a consistent group of agents.When agents have different conceptual models for p,different subgroup of agents know different logic value for p. In this case the n agents are inconsistent in the expression of the logic value for p. The valuation structure of set of agents can be used as a semantic model for the Lattice-valued Logic and fuzzy logic.     

Utopian preference mapping and the utopian preference method for group multiobjective optimization

The individual utopian preference and the group utopian preference on a set of alternatives, and the concept of the utopian preference mapping from the individual utopian preferences, to the group utopian preference, based on the utopian points of the corresponding multiobjective optimization models proposed by decision makers are introduced. Through studying the various fundamental properties of the utopian preference mapping, a method for solving group multiobjective optimization problems with multiple multiobjective optimization models is constructed.     

Design and Implementation of a Storage Virtualization System Based on SCSI Target Simulator in SAN

The ideal storage virtualization system is compatible with all operating systems in storage are anetworks (SANs). However, current storage systems on clustered hosts and multiple operating systems are not practical. This paper presents a storage virtualization system based on a SCSI target simulator in a SAN to solve these problems. This storage virtualization system runs in the target hosts of the SAN, dynamically stores the physical information, and uses the mapping table method to modify the SCSI command addresses. The system uses the bitmap technique to manage the free space. The storage virtualization system provides various functions, such as logical volume resizing, data mirroring, and snapshots, and is compatible with clustered hosts and multiple operating systems, such as Windows NT and RedHat.     

Multi-Dimensional Piece-Wise Self-Affine Fractal Interpolation Model

Iterated function system (IFS) models have been used to represent discrete sequences where the attractor of the IFS is piece-wise self-affine in R 2 or R 3 (R is the set of real numbers). In this paper, the piece-wise self-affine IFS model is extended from R 3 to Rn (n is an integer greater than 3), which is called the multi-dimensional piece-wise self-affine fractal interpolation model. This model uses a "mapping partial derivative" , and a constrained inverse algorithm to identify the model parameters. The model values depend continuously on all the model parameters, and represent most data which are not multi-dimensional self-affine in Rn . Therefore, the result is very general. The class of functions obtained is much more diverse because their values depend continuously on all of the variables, with all the coefficients of the possible multi-dimensional affine maps determining the functions.     

Mata—information generation in distributed information system

The authors discuss the concept of meta-information which is the desription of information system or its subsystems,and proposes algorithms for meta-information generation.Meta-information can be generated in parallel mode and network computation can be used to accelerate meta-information generation.Most existing rough set methods assume information system to be centralized and cannot be applied directly in distributed information system.Data integration ,which is costly,is necessary for such existing methods.However,meta-information integration will eliminate the need of data integration in many cases,since many rough set operations can be done straightforward based on meta-information,and many existing methods can be modified based on meta-information.     

A reference model for database security proxy

How to protect the database, the kernel resources of information warfare, is becoming more and more important since the rapid development of computer and communication technology. As an application-level firewall, database security proxy can successfully repulse attacks originated from outside the network, reduce to zerolevel damage from foreign DBMS products. We enhanced the capability of the COAST‘ s firewall reference model by adding a transmission unit modification function and an attribute value mapping function,describes the schematic and semantic layer reference model, and finally forms a reference model for DBMS security proxy which greatly helps in the design and implementation of database security proxies. This modeling process can clearly separate the system functionality into three layers, define the possible security functions for each layer, and estimate the computational cost for each layer.     

From Matter to Life:Chemistry?Chemistry!

1 Results Animate as well as inanimate matter,living organisms as well as materials,are formed of molecules and of the organized entities resulting from the interaction of molecules with each other.Chemistry provides the bridge between the molecules of inanimate matter and the highly complex molecular architectures and systems which make up living organisms. Synthetic chemistry has developed a very powerful set of methods for constructing ever more complex molecules.Supramolecular chemistry seeks to control the formation of molecular assembly by means of the interactions between the partners.The designed generation of organized architectures requires the handling of information at the molecular level in a sort of molecular programming,thus also linking chemistry with information science.The field of chemistry is the universe of all possible entities and transformations of molecular matter,of which those actually realized in nature represent just one world among all the worlds that await to be created.     

Genetic Algorithms to the Nesting Problem in the Leather Manufacturing Industry

The nesting problem in the leather manufacturing is the problem of placing a set of irregularly shaped pieces (called stencils) on a set of irregularly shaped surfaces (called leathers sheets). This paper presents a novel and promising processing approach. After the profile of leather sheets and stencils is obtained with digitizer, the discretization makes the processing independent of the specific geometrical information. The constraints of profile are regarded thoroughly. A heuristic bottom-left placement strategy is employed to sequentially locate stencils on sheets. The optimal placement sequence and rotation are deterimined by genetic algorithms (GA). A natural concise encoding method is developed to satisfy all the possible requirements of the leather nesting problem. The experimental results show that the proposed algorithm can not only be applied to the normal two-dimensional nesting problem, but also especially suitable for the placement of multiple twodimensional irregular stencils on multiple two-dimensional irregular sheets.     

Fuzzy Set-Based Risk Evaluation Model for Real Estate Projects

With the rapid development of residential real estate market, risk evaluation has been an important task in the process of project. This paper describes a risk evaluation method for residential real estate projects based on fuzzy set theory which uses linguistic variables and respective fuzzy numbers to evaluate the factors. The primary weights of factors and evaluation of alternatives are determined by applying linguistic variables and fuzzy numbers. The notion of Shapley value is used to determine the global value of each factor in accomplishing the overall objective of the risk evaluation process, so the primary weights are revised, thus the importance of factors can be reflected more precisely. A major advantage of the method is that it allows experts and engineers to express their opinions on project risk evaluation in linguistic variables rather than crisp values. An illustration is presented to demonstrate the application of the method in risk evaluation. The results are consistent with the results calculated by conventional risk evaluation method. The research demonstrates that the method is objective and accurate, and is of an application value in the risk evaluation for residential real estate project.     

一种基于Rough集理论的偏好规则获得方法研究

提出一种多准则方案对偏好程度粗度量的一般方法。构造了一种从参考方案集中获取偏好规则并用于全体方案排序和优选的方法。将参考方案对应的偏好关系定量分级表示，应用Rough粗集理论的近似原理对综合偏好关系进行粗近似以获取基于偏好信息的决策规则。计算实例表明新方法能充分表示决策者的偏好，有效避免了LERS系统产生的规则不一致性。     

基于Fuzzy偏好关系的一种方案排序方法

如何根据Ｆｕｚｚｙ偏好关系选择最优方案或进行方案的排序是一个重要的研究内容·针对Ｆｕｚｚｙ偏好关系提出了一种新的方案排序方法·首先描述了Ｆｕｚｚｙ偏好关系及其含义,然后建立了一个最优化模型并通过求解该模型来得到方案的参考排序值,使最终方案的排序结果最大程度地反映决策者的偏好,最后给出了一个算例·     

带有Fuzzy偏好关系的多属性决策方法

针对多属性决策分析中决策者给出关于方案的Fuzzy偏好关系,提出了一种新的多属性决策方法·该方法是通过建立一个线性目标规划模型来确定属性的权重向量并相应地进行方案的排序,得到的方案排序结果体现了决策者给出的方案偏好信息与决策矩阵的集成,并最大程度地反映了决策者的偏好·最后给出了一个算例·     

基于模糊偏好关系及OWA算子的多准则群决策方法

根据一种基于加权的新型模糊偏好关系及计算模糊关系的解析表达式，集结不同决策者的偏好信息，得出群体偏好矩阵，进而求出最优方案．此外，为了使专家在决策时能就所决策的问题达成最大程度的一致，提出了一些“软的”度量专家在决策时所达成的一致程度的一些指标及计算方法，为进一步调整专家决策时的评估信息提供了依据．     

一类不确定信息下的多属性决策问题的研究

对一类不确定信息下的多属性决策问题进行了研究,给出了不确定信息的定义,即在决策过程中决策者不仅不能准确地估计方案的属性值,甚至不能确定同一属性下方案之间的偏好关系.在此基础上,提出使用具有置信结构的偏序关系来表示不确定信息,并给出了解决这类问题的模型和方法.该方法首先采用证据推理算法对不确定信息进行集成,然后使用优势和劣势这2种指数来确定方案间的偏好关系,并在这2种指数的定义中使用了偏序之间的距离公式.最后通过一个例子说明,在信息不确定的情况下该方法能够帮助决策者做出比较合理的决策.     

基于近似随机优势度的随机多属性决策方法

针对属性值为随机变量形式的决策问题,提出了一种随机多属性决策分析方法.首先给出了近似随机优势准则的描述及相关分析,并提出了近似随机优势度的定义和计算公式;然后,依据近似随机优势准则判断并确定两两方案之间比较所具有的优势关系,进而计算得到方案之间的近似随机优势度,并构建相应的近似随机优势度矩阵.进一步地,通过计算可以得到每个方案的总体相对近似随机优势程度,从而得到所有方案的排序结果.最后,通过一个算例说明了所给出方法的可行性和有效性.     

置换群与群决策的排序数估计

在决策方案优劣等级排序分析和多层序等价类划分的基础上,构造了等级排序置换群,并提出了群决策中序位相同的可能排序数的解析和递推估计式.分析结果显示,注重维护决策方案的齐次性和协调决策者的意见冲突是提高群偏好集结有效性的重要方法.     

偏好关系的对偶性

建立了基本二元关系以及偏好关系的对偶理论．引入并讨论了二元关系之间的基本性质以及偏好关系之间的一些性质的对偶性．建立了基本二元关系以及偏好关系的各种性质之间的蕴涵式．定义了一些新的序关系，并对各种偏序关系所满足的基本性质给出了完整的刻划．     

群决策中基于不同偏好信息的相对熵集成方法

研究了基于不同形式偏好信息的群决策问题.在群决策中专家根据个人的偏好,常常对决策方案集给出4种不同形式的偏好信息,包括直接反映决策方案优劣次序的序关系值和效用值,以及2个决策方案成对比较互反判断矩阵和模糊互补判断矩阵.首先给出了序关系值、互反判断矩阵和模糊互补判断矩阵3种偏好信息均转化为效用值形式的计算公式, 然后从相对熵的概念出发,提出了一种相对熵最优化的集成模型,给出了模型的解.最后进行了2个实例分析,结果表明所提出集成方法是有效的.     

集值信息系统中的粗糙集扩展模型

根据属性值的相容程度在集值信息系统上定义了一种新的关系--α相容关系;给出了相应的近似集合和相关性质,并讨论了它与相容关系、拟序关系之间关系;提出了基于这种关系的粗糙集扩展模型,并分析比较了该模型与现有的粗糙集扩展模型之间的关系;最后,通过一个实例来直观地解释这些关系之间的联系.