Likelihood ratio estimation in forensic identification using similarity and rarity

Forensic identification is the task of determining whether or not observed evidence arose from a known source. It is useful to associate probabilities with identification/exclusion opinions, either for presentation in court or to evaluate the discriminative power of a given set of attributes. At present, in most forensic domains outside of DNA evidence, it is not possible to make such a statement since the necessary probability distributions cannot be computed with reasonable accuracy, although the probabilistic approach itself is well-understood. In principle, it involves determining a likelihood ratio (LR) – the ratio of the joint probability of the evidence and source under the identification hypothesis (that the evidence came from the source) and under the exclusion hypothesis (that the evidence did not arise from the source). Evaluating the joint probability is computationally intractable when the number of variables is even moderately large. It is also statistically infeasible since the number of parameters to be determined from the data is exponential with the number of variables. An approximate method is to replace the joint probability by another probability: that of distance (or similarity) between evidence and object under the two hypotheses. While this reduces to linear complexity with the number of variables, it is an oversimplification leading to errors. We consider a third method which decomposes the LR into a product of two factors, one based on distance and the other on rarity. This result, which is exact for the univariate Gaussian case, has an intuitive appeal – forensic examiners assign higher importance to rare feature values in the evidence and low importance to common feature values. We generalize this approach to more complex data such as vectors and graphs, which makes LR estimation computationally tractable. Empirical evaluations of the three methods, done with several data types (continuous features, binary features, multinomial and graph) and several modalities (handwriting with binary features, handwriting with multinomial features and footwear impressions with continuous features), show that the distance and rarity method is significantly better than the distance only method.     

Distance learning for similarity estimation

In this paper, we present a general guideline to find a better distance measure for similarity estimation based on statistical analysis of distribution models and distance functions. A new set of distance measures are derived from the harmonic distance, the geometric distance, and their generalized variants according to the Maximum Likelihood theory. These measures can provide a more accurate feature model than the classical Euclidean and Manhattan distances. We also find that the feature elements are often from heterogeneous sources that may have different influence on similarity estimation. Therefore, the assumption of single isotropic distribution model is often inappropriate. To alleviate this problem, we use a boosted distance measure framework that finds multiple distance measures which fit the distribution of selected feature elements best for accurate similarity estimation. The new distance measures for similarity estimation are tested on two applications: stereo matching and motion tracking in video sequences. The performance of boosted distance measure is further evaluated on several benchmark data sets from the UCI repository and two image retrieval applications. In all the experiments, robust results are obtained based on the proposed methods.     

Semantic similarity estimation from multiple ontologies

The estimation of semantic similarity between words is an important task in many language related applications. In the past, several approaches to assess similarity by evaluating the knowledge modelled in an ontology have been proposed. However, in many domains, knowledge is dispersed through several partial and/or overlapping ontologies. Because most previous works on semantic similarity only support a unique input ontology, we propose a method to enable similarity estimation across multiple ontologies. Our method identifies different cases according to which ontology/ies input terms belong. We propose several heuristics to deal with each case, aiming to solve missing values, when partial knowledge is available, and to capture the strongest semantic evidence that results in the most accurate similarity assessment, when dealing with overlapping knowledge. We evaluate and compare our method using several general purpose and biomedical benchmarks of word pairs whose similarity has been assessed by human experts, and several general purpose (WordNet) and biomedical ontologies (SNOMED CT and MeSH). Results show that our method is able to improve the accuracy of similarity estimation in comparison to single ontology approaches and against state of the art related works in multi-ontology similarity assessment.     

Efficient model similarity estimation with robust hashing

Software and Systems Modeling - As model-driven engineering (MDE) is increasingly adopted in complex industrial scenarios, modeling artefacts become a key and strategic asset for companies. As...     

Statistical estimation of multiple measures of similarity

In this paper, statistical estimates for linear-fractional multiple measures of similarity of the K(T, C _Δ)-type are considered. Examples of computing multiple similarity measures, their standard errors, and confidence intervals are presented.     

Distributed and robust parameter estimation of IIR systems using incremental particle swarm optimization

In recent years because of substantial use of wireless sensor network the distributed estimation has attracted the attention of many researchers. Two popular learning algorithms: incremental least mean square (ILMS) and diffusion least mean square (DLMS) have been reported for distributed estimation using the data collected from sensor nodes. But these algorithms, being derivative based, have a tendency of providing local minima solution particularly for minimization of multimodal cost function. Hence for problems like distributed parameters estimation of IIR systems, alternative distributed algorithms are required to be developed. Keeping this in view the present paper proposes two population based incremental particle swarm optimization (IPSO) algorithms for estimation of parameters of noisy IIR systems. But the proposed IPSO algorithms provide poor performance when the measured data is contaminated with outliers in the training samples. To alleviate this problem the paper has proposed a robust distributed algorithm (RDIPSO) for IIR system identification task. The simulation results of benchmark IIR systems demonstrate that the proposed algorithms provide excellent identification performance in all cases even when the training samples are contaminated with outliers.     

Deriving similarity for Semantic Web using similarity graph

One important research challenge of current Semantic Web is resolving the interoperability issue across ontologies. The issue is directly related to identifying semantics of resources residing in different domain ontologies. That is, the semantics of a concept in an ontology differs from others according to the modeling style and intuition of the knowledge expert even though they are the same forms of a concept in each respective ontology. In this paper, we propose a similarity measure to resolve the interoperability issue by using a similarity graph. The strong point of this paper is that we provide a precise mapping technique and similarity properties to derive the similarity. The novel contribution of this paper is that we provide a core technique of computing similarity across ontologies of Semantic Web. This research was supported by the MIC (Ministry of Information and Communication), Korea, under the ITRC (Information Technology Research Center) support program supervised by the IITA (Institute of Information Technology Assessment).     

Distributed estimation algorithms for nonlinear systems

In this paper, we consider the problem of combining the local conditional distributions of a random variable which have been generated by local observers having access to their private information. Sufficient statistics for the local distributions are communicated to a coordinator, who attempts to reconstruct the global centralized distribution using only the communicated statistics. We obtain a distributed processing algorithm which recovers exactly the centralized conditional distribution. The results can be applied in designing distributed hypothesis-testing algorithms for event-driven systems.     

A robust non-iterative method for similarity transform estimation

A non-iterative and robust method—direct outliers remove (DOR) is proposed, which efficiently estimates the similarity transform based on a data set containing both correct and incorrect correspondences. Unlike hypothesize-and-test methods such as Random Sample Consensus algorithm and its variants, DOR removes mismatches by exploring all the correspondences only once, using two invariant features of similarity transform. One is the angles between two vectors and the other is the length ratios of corresponding vectors. Given two images related by similarity transform, experiments demonstrate that all the mismatches introduced in matching stage could be detected and removed. Without losing computational accuracy, DOR is faster compared with several hypothesize-and-test algorithms, especially when the percentage of correct correspondence is relatively low.     

基于医疗本体的语义相似度评估方法

为了能够正确地理解医疗概念和精确地分析临床记录,提出了一种基于概念信息量的方法来衡量概念之间的语义相似度.引进了计算概念信息量的算法,从医疗本体的分类知识中来计算概念的信息量.介绍和分析了常用的语义相似度算法,根据概念的信息量来重定义这些语义相似度算法,产生新的基于概念信息量的语义相似度算法.通过使用一个医疗术语的评估标准和一个标准的医疗本体来评估和比较这些算法.实验结果表明,相比常用的语义相似度算法,重定义后的算法有效地改善了概念相似性评估的准确性.     

Pose estimation by fusing noisy data of different dimensions

A method for fusing and integrating different 2D and 3D measurements for pose estimation is proposed. The 2D measured data is viewed as 3D data with infinite uncertainty in particular directions. The method is implemented using Kalman filtering. It is robust and easily parallelizable     

分布式奇异值分解最小平方估计算法

对奇异值（SVD）分解求解最小平方估计的问题进行了研究。提出迭代式分割与合并的算法（IDMSVD）,目的是改善奇异值分解在估计参数时非常耗费时间以及内存空间的问题。基于IDMSVD提出了分布式迭代式分割与合并算法（MRDSVD）,使用Hadoop平台的MapReduce来实现,实验结果显示,IDMSVD可以有效改善SVD求最小平方解耗费运行时间与内存空间的问题,MRDSVD算法可进一步改善IDMSVD的运行时间。     

Distributed parameter and state estimation in petroleum reservoirs

In this paper, we describe some of the key elements of the data assimilation problem for multiphase flow in petroleum reservoirs that make the problem distinctly different from data assimilation problems in weather or oceanography. Most importantly, the reservoir is often initially in a state of static equilibrium, the number of model parameters may be greater than the number of state variables, and the evolution of some of the state variables proceed monotonically from the initial state (low water saturation) to a final state (high water saturation). As a result of the differences, data assimilation is sometimes applied with a focus on estimation of model parameters.     

Joint structural similarity and entropy estimation for coded-exposure image restoration

Multimedia Tools and Applications - We address the image deblurring using coded exposure which can keep image content that may be lost by a traditional shutter. In the restoration of a coded...     

Volume analysis using multimodal surface similarity

The combination of volume data acquired by multiple modalities has been recognized as an important but challenging task. Modalities often differ in the structures they can delineate and their joint information can be used to extend the classification space. However, they frequently exhibit differing types of artifacts which makes the process of exploiting the additional information non-trivial. In this paper, we present a framework based on an information-theoretic measure of isosurface similarity between different modalities to overcome these problems. The resulting similarity space provides a concise overview of the differences between the two modalities, and also serves as the basis for an improved selection of features. Multimodal classification is expressed in terms of similarities and dissimilarities between the isosurfaces of individual modalities, instead of data value combinations. We demonstrate that our approach can be used to robustly extract features in applications such as dual energy computed tomography of parts in industrial manufacturing.     

Marking student programs using graph similarity

We present a novel approach to the automated marking of student programming assignments. Our technique quantifies the structural similarity between unmarked student submissions and marked solutions, and is the basis by which we assign marks. This is accomplished through an efficient novel graph similarity measure (AssignSim). Our experiments show good correlation of assigned marks with that of a human marker.     

Image forgery detection using image similarity

Multimedia Tools and Applications - Ideally, sophisticated image forgery methods leave no perceptible evidence of tampering. In response to such stringent context, researchers have proposed digital...