基于Pignistic相似度与改进DS证据理论的综合评估方法

针对标准的Dempster-Shafer（DS）证据理论在证据冲突情况下决策准确率低与评估稳定性差的问题,提出一种基于Pignistic相似度与改进DS证据理论的综合评估方法。由Pignistic概率转换求得各类证据的相似度,计算对应的支持度与自适应权重,引入不确定命题推导出改进的证据合成规则。结合某航道实际通航状况进行算例分析,验证该方法的有效性与可靠性。分析结果表明,所提出的方法能正确处理不确定性条件下证据之间的冲突并能给出稳定决策结果。     

基于交补集和Pignistic变换的证据组合方法

针对D-S证据组合公式及其改进公式的局限性,提出一种基于交补集权重和Pignistic概率变换的改进组合方法。基于交补集理论推导出新的基本概率分配函数,对交补集权重因子进行量化,得到基于交补集权重的证据组合公式,利用Pignistic概率变换法对已获得的各命题的信度值进行重新分配,以降低组合顺序对合成结果的影响,同时可获得更可靠的决策依据。实例分析结果表明,与其他改进方法相比,该组合方法在解决冲突证据、一票否决、鲁棒性、公平性和决策有效性等方面均有明显的优势。     

基于不确定度的基本概率赋值概率转换方法

基本概率赋值概率转换是合成证据用于辅助决策的主要方法之一。针对现有转换方法中存在的缺少合理转换依据,灵活性和客观性不能兼顾的问题,提出一种基于不确定度加权的概率转换方法。该方法将概率转换视为一种决策过程,选择原BPA与最特异BPA间的Jousselme距离作为不确定性量度,通过“假设-校验”的方法联合求解最特异BPA与转换概率。实验算例表明,所提出方法的处理过程与人们的决策过程一致,概率转换结果合理有效。     

一种基于“交并集”和Pignistic概率的证据冲突的改进方法

针对D-S证据理论无法解决高冲突证据的缺陷，通过对现有几种证据冲突的改进方法进行分析，提出了基于“交并集”和Pignistic概率的改进方法。本文放宽D-S组合规则的假设，只要求证据在组合时至少有一条是真实的，如果证据A和B相互支持，说明它们都是真实的，可以用“交集”运算将证据的信度聚焦在它们的芟集上；如果证据A和B相互冲突，表明不知道哪一条证据是真实的，则用“并集”运算将信度聚焦在它们的并集上，即证据支持A或B中的一个，这种思路更符合人类的直觉。由于在目标识别系统中，最终决策是单个待识目标，因此以还要用Pignistic概率转换法将多元素命题的BPA再分配给它的各个组合元素，最后，信度最高的元素作为结果进行输出。实验表明，本文方法在解决证据冲突方面较其他方法拥有明显的优势。使用本文方法时，证据的融合顺序对融合结果没有影响，因此可以很方便地编程实现。     

基于信息无损视角的DS/AHP群决策方法

因参考点选择不恰当及折扣方式不合理,DS/AHP 群决策方法存在决策信息损失的问题。为此,基于由决策主体推理判断出的互斥方案组和七级标度陒对偏好信息,构建能够对所有决策主体在特定属性上进行偏好集成的主体信息融合模型。在此基础上参照传统方法中由知识矩阵向BPA函数转换的思想,以及利用Dempster组合规则进行信息融合的思想,提出能够综合集成所有属性上证据信息的方法步骤,并通过案例模拟验证该方法的科学有效性和应用可行性,其 Pignistic概率与标准结果之间的总差异程度较小。     

基于概率转换函数与邓熵的关联分类融合算法

在信息高度冲突情况下，基于信度转移模型的关联二分类算法会出现分类精确率低、计算量大等问题。为解决这些问题，提出一种关联分类融合算法。由类关联规则的置信度构造基本概率指派函数，给出分类规则；基于Pignistic概率转换函数重构邓熵，结合PPT散度优化分类规则；通过解决概率转换函数的高冲突，使证据的高冲突问题得到有效改善，实验结果验证了这一点。实验结果还表明，分类准确性得到有效提升，与BJS散度等信度转移模型相比，计算量显著降低。     

D-S证据理论决策规则分析

该文对D-S证据理论的多种决策规则进行了详尽的分析比较,并提出两种新的基于pignistic概率的决策规则。实验表明,提出的决策规则能获得更高的性能评价,并可以有效利用现有信息扩大各命题之间的信度差异,在不提高决策风险的前提下,使满足决策阈值的待选目标数降至最少,更有利于系统的最终决策。对D-S证据理论决策规则的研究结果同样适用于其它基于信度的不确定推理方法。     

Markov决策过程的蚁群规划算法

在智能规划问题上,寻找规划解都是NP甚至NP完全问题,如果动作的执行效果带有不确定性,如在Markov决策过程的规划问题中,规划的求解将会更加困难,现有的Markov决策过程的规划算法往往用一个整体状态节点来描述某个动作的实际执行效果,试图回避状态内部的复杂性,而现实中的大量动作往往都会产生多个命题效果,对应多个命题节点。为了能够处理和解决这个问题,提出了映像动作,映像路节和映像规划图等概念,并在其基础上提出了Markov决策过程的蚁群规划算法,从而解决了这一问题。并且证明了算法得到的解,即使在不确定的执行环境下,也具有不低于一定概率的可靠性。     

一种基于Pignistic概率距离的合成公式

针对多证据源信息融合过程中证据源间存在的冲突问题,提出了一种基于Pignistic概率距离的合成公式。利用Pignistic概率距离构造证据可信度;再利用证据可信度修正证据体,以改进合成公式;利用改进的合成公式对证据源进行融合。算例结果表明,改进合成公式的融合结果合理有效,与其他方法相比有更好的适用性、可靠性和较快的运算速度。     

同Vague谓词命题的概率逻辑

基于对Vague（或Fuzzy）概念的一种新的认知,使用随机集和概率论,引入了论域表达式及其适当测度的概念。进一步地,通过引入同Vague谓词命题及其他们真的概率（简称概率真度）的概念,提出了一种新的非经典命题逻辑,称为同Vague谓词命题的概率逻辑,并提供了其逻辑规律。其逻辑规律表明它与经典逻辑具有优良的和谐性。比较了同Vague谓词命题的概率逻辑与模糊逻辑的长处与不足。     

Interaction transform for bi-set functions over a finite set

Set functions appear as a useful tool in many areas of decision making and operations research, and several linear invertible transformations have been introduced for set functions, such as the Möbius transform and the interaction transform. The present paper establish similar transforms and their relationships for bi-set functions, i.e. functions of two disjoint subsets. Bi-set functions have been recently introduced in decision making (bi-capacities) and game theory (bi-cooperative games), and appear to open new areas in these fields.     

New sequential exact Euclidean distance transform algorithms based on convex analysis

We present several sequential exact Euclidean distance transform algorithms. The algorithms are based on fundamental transforms of convex analysis: The Legendre Conjugate or Legendre–Fenchel transform, and the Moreau envelope or Moreau-Yosida approximate. They combine the separability of the Euclidean distance with convex properties to achieve an optimal linear-time complexity.We compare them with a Parabolic Envelope distance transform, and provide several extensions. All the algorithms presented perform equally well in higher dimensions. They can naturally handle grayscale images, and their principles are generic enough to apply to other transforms.     

Target identification based on the transferable belief model interpretation of dempster-shafer model

This paper explains how multisensor data fusion and target identification can be performed within the transferable belief model (TBM), a model for the representation of quantified uncertainty based on belief functions. We present the underlying theory, in particular the general Bayesian theorem needed to transform likelihoods into beliefs and the pignistic transformation needed to build the probability measure required for decision making. We present how this method applies in practice. We compare its solution with the classical one, illustrating it with an embarrassing example, where the TBM and the probability solutions completely disagree. Computational efficiency of the belief-function solution was supposedly proved in a study that we reproduce and we show that in fact the opposite conclusions hold. The results presented here can be extended directly to many problems of data fusion and diagnosis.     

Evaluation of discrete transforms for use in digital speech recognition

Traditionally FFT (fast Fourier transform) has been utilized in recognition algorithms involving speech. Other discrete transforms such as Walsh-Hadamard transform (WHT) and rapid transform (RT) can play equally important roles in the recognition process as they have advantages in implementation and hardware realization. The capability of these transforms in recognizing phonemes based on training matrices and various matching criteria is investigated. The speech data base consists of ten sentences spoken by ten different speakers (all male). For recognition purposes the speech is sectioned into 10 ms intervals and is sampled at 20 KHz. Training matrices for all the three transforms are developed. Test matrices in the transform domain are compared with the prototypes based on these criteria which led to the decision process. WHT and RT appear to offer promise and potential compared to FFT as the former are easier to implement and as the yield recognition results comparable to those of the FFT. Other distance measures and recognition schemes are proposed for improving the classification accuracy.     

Computing distributions and moments in polling models by numerical transform inversion

We show that probability distributions and moments of performance measures in many polling models can be effectively computed by numerically inverting transforms (generating functions and Laplace transforms). We develop new efficient iterative algorithms for computing the transform values and then use our recently developed variant of the Fourier-series method to perform the inversion. We also show how to use this approach to compute moments and asymptotic parameters of the distributions. We compute a two-term asymptotic expansion of the tail probabilities, which turns out to be remarkably accurate for small tail probabilities. The tail probabilities are especially helpful in understanding the performance of different polling disciplines. For instance, it is known that the exhaustive discipline produces smaller mean steady-state waiting times than the gated discipline, but we show that the reverse tends to be true for small tail probabilities. The algorithms apply to describe the transient behavior of stationary or non-stationary models as well as the steady-state behavior of stationary models. We demonstrate effectiveness by analyzing the computational complexity and by doing several numerical examples for the gated and exhaustive service disciplines, with both zero and non-zero switchover times. We also show that our approach applies to other polling models. Our main focus is on computing exact tail probabilities and asymptotic approximations to them, which seems not to have been done before. However, even for mean waiting times, our algorithm is faster than previous algorithms for large models. The computational complexity of our algorithm is O(N) for computing performance measures at one queue and O(N¹⁺) for computing performance measures at all queues, where N is the number of queues and is typically between 0.6 and 0.8.     

基于Log-Gabor小波变换和证据推理的车型识别*

针对车型识别问题,提出了一种基于特征车的车型识别方法——基于Log-Gabor小波变换和DS证据推理的车型识别算法。该算法先对特殊车辆图像进行多分辨率的Log-Gabor小波变换,最后形成车辆Log-Gabor特征。将1-a-1多分类SVM应用于基本概率分配函数的确定,使用证据推理的方法得到车型识别的结果。实验结果表明该方法是有效、可行的。