Dissimilarity Measure Research Articles

A generalized belief dissimilarity measure based on weighted conflict belief and distance metric and its application in multi-source data fusion

Dissimilarity measure between basic probability assignments (BPAs) in the Dempster-Shafer evidence structure is a vibrant research topic in artificial intelligence. However, there are flaws in the existing measurements. In particular, it is insufficient to characterize dissimilarity only from either evidential distance or conflict belief for a BPA. As such, we propose a new dissimilarity measure which takes into consideration both distance measure and conflict belief among betting commitments. These two factors complement with each other. Distance measure reflects diversity between the focal elements of two pieces of evidence. That is to say, the more intersections between the bodies of evidence (BOE) of two data sources, the more reliable it acts as a dissimilarity measure. Conversely, the conflict belief which is created based on the transformed Pignistic probability characterizes the product of singleton's belief from two pieces of evidence whose intersection is empty. It quantifies dissimilarity measure more efficiently when the focal elements of two pieces of evidence have small intersect. Theoretically, the new dissimilarity measure satisfies reflexivity, symmetry, nonnegativity, nondegeneracy and some other properties. Comparative analysis is provided with some cases to demonstrate the applicability and validity of the proposed dissimilarity measure. To determine the weight and reliability of evidence, the new dissimilarity measure among evidence and uncertainty of BPA are used. The dissimilarity metric is further applied for multi-source data fusion together with uncertainty measure of belief structure. The application of large-scale group decision making (LSGDM) problem is given to illustrate the effectiveness of the proposed multi-source data fusion process.

A spatially‐weighted AMH copula‐based dissimilarity measure for clustering variables: An application to urban thermal efficiency

AbstractInvestigating thermal energy demand is crucial for developing sustainable cities and the efficient use of renewable sources. Despite the advances made in this field, the analysis of energy data provided by smart grids is currently a demanding challenge due to their complex multivariate structure and high dimensionality. In this article, we propose a novel copula‐based dissimilarity measure suitable for analyzing district heating demand and introduce a procedure to apply it to high‐temporal resolution panel data. Inspired by the characteristics of the considered data, we explore the usefulness of the Ali‐Mikhail‐Haq copula in defining a new dissimilarity measure to cluster variables in the hierarchical framework. We show that our proposal is particularly sensitive to small dissimilarities based on tiny differences in the strength of the dependence between the involved random variables. Therefore, the measure we introduce is able to distinguish between objects with low dissimilarity better than standard rank‐based dissimilarity measures. Moreover, our proposal considers a weighted version of the copula‐based dissimilarity that embeds the spatial location of the involved objects. We investigate the proposed measure through Monte Carlo studies and compare it with an analogous dissimilarity measure based on Kendall's correlation. Finally, the application to real data concerning the Italian city Bozen‐Bolzano makes it possible to find clusters of buildings homogeneous with respect to their main characteristics, such as energy efficiency and heating surface. In turn, our findings may support the design, expansion, and management of district heating systems.

Urban consumption dissimilarities and subjective wellbeing: Evidence from mobile big data in China

The inequality and spatial unevenness in the distribution of economic resources can both significantly impact individual wellbeing. Data on residents’ income at a fine spatial scale are often not available in developing countries. In this study, we utilize a novel dataset derived from mobile big data with high spatial resolutions to estimate residents’ income at the district level in China. By combining this measure of consumption dissimilarity with household survey data, our findings reveal that even after controlling for overall levels of inequality, the concentration of high-spending households within a district has a negative effect on residents’ life satisfaction while the concentration of low-spending households has an opposite effect. These results contribute to the emerging research on understanding the spatial dimensions of economic inequalities.

A novel clustering method for complex signals and feature extraction based on advanced information-based dissimilarity measure

In this paper, a new dissimilarity measure for more accurate feature extraction and clustering is put forward. The method is proposed from the perspective of the weighted-probability distribution of dispersion patterns and their rank order statistics, where an effective quantization procedure is provided and the loss of information can be reduced. The proposed dissimilarity is applied in the multidimensional scaling (MDS) method to investigate simulated and reality-based signals. The comparative experiment shows that the clustering results of the proposed technique are clearer and more appropriate. State-of-the-art techniques and conventional methods are both included in the comparative experiments. In particular, for the heartbeat signals, it is discovered that the distribution of the weighted-probabilities of the dispersion patterns can discriminate subjects with different physiological conditions and exhibit visible changes of the subject’s dynamical features when aging and disease attacks are taken place, which can be regarded as a microscopic insight of the dynamical mechanisms.

An approximate tensor singular value decomposition approach for the fast grouping of whole-body human poses

We propose a fast method to group similar human poses from single-view depth maps using an approximate singular value decomposition approach in the tensor domain. To this end, the input tensor is decomposed, and a representation tensor is learned using the tensor-QR decomposition and ℓ2,1 norm minimization. The spatial structure and, thereby, the spatial information in each depth map, vital in performing accurate grouping of human poses, is preserved by adopting the 3D tensor approach in treating the data. For the seamless integration of discriminatory information of each pose, a new dissimilarity measure based on sub-tensors is devised and integrated into the optimization problem. Experimental analysis showcases the ability of the proposed method to achieve 10%–15% improvement in the grouping results compared with the state-of-the-art counterparts. The proposed algorithm also converges in less than ten iterations on average, thereby improving CPU time.

Fractal-based basic probability assignment: A transient mass function

Effectively measuring the information content of Basic Probability Assignments (BPA) is a crucial prerequisite for addressing uncertain information within the framework of Dempster-Shafer theory. In previous research, uncertainty and dissimilarity measures for BPAs were often treated separately to counteract the impact of incomplete mutual exclusion between focal elements. In this paper, we introduce a Transient Mass Function (TMF) based on the process of pignistic probability transformation to forge a link between probabilistic and evidential information. As a transient state of BPA, the TMF not only considers the structural features of the focal element but also maintains its numerical content. Carrying out probabilistic information measures on the TMF is equivalent to performing belief information measures on the original BPA. Finally, validated by numerical simulations, TMF provides a unified and effective metric framework for evidential information.

Online Deterministic Annealing for Classification and Clustering.

Inherent in virtually every iterative machine learning algorithm is the problem of hyperparameter tuning, which includes three major design parameters: 1) the complexity of the model, e.g., the number of neurons in a neural network; 2) the initial conditions, which heavily affect the behavior of the algorithm; and 3) the dissimilarity measure used to quantify its performance. We introduce an online prototype-based learning algorithm that can be viewed as a progressively growing competitive-learning neural network architecture for classification and clustering. The learning rule of the proposed approach is formulated as an online gradient-free stochastic approximation algorithm that solves a sequence of appropriately defined optimization problems, simulating an annealing process. The annealing nature of the algorithm contributes to avoiding poor local minima, offers robustness with respect to the initial conditions, and provides a means to progressively increase the complexity of the learning model, through an intuitive bifurcation phenomenon. The proposed approach is interpretable, requires minimal hyperparameter tuning, and allows online control over the performance-complexity tradeoff. Finally, we show that Bregman divergences appear naturally as a family of dissimilarity measures that play a central role in both the performance and the computational complexity of the learning algorithm.

Time-varying multi-period emergency decision-making based on dissimilarity measure and definite integral in the picture fuzzy context

Time-varying multi-period emergency decision-making based on dissimilarity measure and definite integral in the picture fuzzy context

Clustering Analysis of Multivariate Data: A Weighted Spatial Ranks-Based Approach

Determining the right number of clusters without any prior information about their numbers is a core problem in cluster analysis. In this paper, we propose a nonparametric clustering method based on different weighted spatial rank (WSR) functions. The main idea behind WSR is to define a dissimilarity measure locally based on a localized version of multivariate ranks. We consider a nonparametric Gaussian kernel weights function. We compare the performance of the method with other standard techniques and assess its misclassification rate. The method is completely data-driven, robust against distributional assumptions, and accurate for the purpose of intuitive visualization and can be used both to determine the number of clusters and assign each observation to its cluster.

A guide to between‐community functional dissimilarity measures

One of the effective tools to study the variation between communities is the use of pairwise dissimilarity indices. Besides species as variables, the involvement of trait information provides valuable insight into the functioning of ecosystems. In recent years, a variety of indices have been proposed to quantify functional dissimilarity between communities. These indices follow different approaches to account for between‐species similarities in calculating community dissimilarity, yet they all have been proposed as straightforward tools. In this paper, we review the trait‐based dissimilarity indices available in the literature and identify the most important conceptual and technical properties that differentiate among them, and that must be considered before their application. We identify two primary aspects that need to be considered before choosing a functional dissimilarity index. The first one is the way communities are represented in the trait space. The three main types of representations are the typical values, the discrete sets using the combination of species × sites and species × traits matrices, and the hypervolumes. The second decision is the concept of dissimilarity to follow, including two options: distances and disagreements. We use the above scheme to discuss the available functional dissimilarity indices and evaluate their relations to each other, their capabilities, and accessibility.

Metrics for network comparison using egonet feature distributions

Identifying networks with similar characteristics in a given ensemble, or detecting pattern discontinuities in a temporal sequence of networks, are two examples of tasks that require an effective metric capable of quantifying network (dis)similarity. Here we propose a method based on a global portrait of graph properties built by processing local nodes features. More precisely, a set of dissimilarity measures is defined by elaborating the distributions, over the network, of a few egonet features, namely the degree, the clustering coefficient, and the egonet persistence. The method, which does not require the alignment of the two networks being compared, exploits the statistics of the three features to define one- or multi-dimensional distribution functions, which are then compared to define a distance between the networks. The effectiveness of the method is evaluated using a standard classification test, i.e., recognizing the graphs originating from the same synthetic model. Overall, the proposed distances have performances comparable to the best state-of-the-art techniques (graphlet-based methods) with similar computational requirements. Given its simplicity and flexibility, the method is proposed as a viable approach for network comparison tasks.

Metric Learning as a Service With Covariance Embedding

Metric learning as a service (MLaaS) represents one of the main learning streams to handle complex datasets in service computing research communities and industries. A common approach for dealing with high-dimensional and complex datasets is employing a feature embedding algorithm to compress data through dimension reduction while optimizing intra-class distance. To create generalizable MLaaS for high-performance artificial intelligence applications with high-dimensional Big Data, a robust and meaningful embedding space representation by efficiently optimizing both intra-class and inter-class relationships is required. We developed a novel MLaaS methodology that incorporates covariance to signify the direction of the linear relationship between data points in an embedding space. Our covariance-based feature embedding architecture introduces three different yet complementary mapping functions: inner-class mapping, intra-class with semi-inter-class mapping, and intra- and inter-class mapping. Unlike conventional metric learning, our covariance-embedding-enhanced approach is more expressive and explainable for computing similar or dissimilar measures and can capture positive, negative, or neutral relationships. Our MLaaS framework ensures efficient, composable, and extensible metric learning by supporting the selection of dimension reduction and data compression methods. Experiments conducted using various benchmark datasets demonstrate that the proposed model can obtain higher-quality, more separable, and more expressive embedding representations than existing models.

ANALISIS GEROMBOL PADA DATA DERET WAKTU PENDERITA COVID-19 PROVINSI JAWA BARAT

This study aims to apply clustering techniques to time series data. Time series models can be formed for all research data objects, so many research objects need to be grouped so that the resulting model becomes more efficient. The object used in this study was data on Covid-19 sufferers from 27 regencies and cities in West Java Province. All objects were analyzed by time series to produce 27 models. All objects' data patterns and models have many similarities, so clustering can be done. Clustering models using the Ward method and the Piccolo dissimilarity measure. The optimum cluster uses the Hartigan and Ball indices to obtain 3 clusters

On the issue of convergence of certain divergence measures related to finding most nearly compatible probability distribution under the discrete set-up

In the context of finding most nearly compatible probability distributions under the discrete set-up, the role(s) of divergence measures as pseudo-distance (equivalently as measures of dissimilarity) measures are of paramount importance. For a detailed discussion on various measures of statistical divergence and its properties, see Pardo (2018) and the references cited therein. Recently, Ghosh and Balakrishnan (2015) have utilized several of such divergence measures, such as the Power divergence, and several other measures. However, in search for the most nearly compatible (or incompatible) distributions, the convergence of the iterative algorithms remains a challenging issue. In this article, we put forward a sketch of the proof regarding the convergence of the iterative algorithm(s) for certain divergence measures under some mild conditions. The proof for the more general case still remains an open problem.

SpikeShip: A method for fast, unsupervised discovery of high-dimensional neural spiking patterns.

Neural coding and memory formation depend on temporal spiking sequences that span high-dimensional neural ensembles. The unsupervised discovery and characterization of these spiking sequences requires a suitable dissimilarity measure to spiking patterns, which can then be used for clustering and decoding. Here, we present a new dissimilarity measure based on optimal transport theory called SpikeShip, which compares multi-neuron spiking patterns based on all the relative spike-timing relationships among neurons. SpikeShip computes the optimal transport cost to make all the relative spike-timing relationships (across neurons) identical between two spiking patterns. We show that this transport cost can be decomposed into a temporal rigid translation term, which captures global latency shifts, and a vector of neuron-specific transport flows, which reflect inter-neuronal spike timing differences. SpikeShip can be effectively computed for high-dimensional neuronal ensembles, has a low (linear) computational cost that has the same order as the spike count, and is sensitive to higher-order correlations. Furthermore, SpikeShip is binless, can handle any form of spike time distributions, is not affected by firing rate fluctuations, can detect patterns with a low signal-to-noise ratio, and can be effectively combined with a sliding window approach. We compare the advantages and differences between SpikeShip and other measures like SPIKE and Victor-Purpura distance. We applied SpikeShip to large-scale Neuropixel recordings during spontaneous activity and visual encoding. We show that high-dimensional spiking sequences detected via SpikeShip reliably distinguish between different natural images and different behavioral states. These spiking sequences carried complementary information to conventional firing rate codes. SpikeShip opens new avenues for studying neural coding and memory consolidation by rapid and unsupervised detection of temporal spiking patterns in high-dimensional neural ensembles.

A dynamic emergency response decision-making method considering the scenario evolution of maritime emergencies

In order to reduce the losses caused by maritime emergencies, it is crucial to select an alternative for timely response after the occurrence of maritime emergencies. However, maritime emergencies involve uncertainty, complexity, and information incompleteness within the scenario evolution process, making it difficult for decision-makers to evaluate with crisp numbers in a completely rational situation. Moreover, decision-makers need to dynamically adjust the emergency response based on the changing scenario. In response, this paper proposes a novel dynamic emergency response decision-making method that not only deals with the problem of fuzzy decision-making under bounded rationality, but also ameliorates the effects of early evaluations within the dynamic adjustment process on later decisions. Specifically, the decision-makers’ evaluations of alternatives in a hesitant fuzzy environment are represented as probabilistic linguistic term sets (PLTSs). Subsequently, a new dissimilarity measure for the PLTSs based on the Jensen-Shannon divergence is proposed and combined with prospect theory (PT), which describes the decision-makers’ psychological behavior. From this, a new static rating calculation method is developed. Furthermore, a static rating-based, dynamic rating method that can reflect the historical evolution of alternatives is put forward to rank alternatives in order to determine the best response option. Finally, an example of a maritime accident is applied to illustrate the feasibility of the proposed method, and a comparative analysis is performed to validate its effectiveness as a decision-making support tool for maritime emergency managers.

Self-Supervised Approach to Addressing Zero-Shot Learning Problem

In recent years, self-supervised learning has had significant success in applications involving computer vision and natural language processing. The type of pretext task is important to this boost in performance. One common pretext task is the measure of similarity and dissimilarity between pairs of images. In this scenario, the two images that make up the negative pair are visibly different to humans. However, in entomology, species are nearly indistinguishable and thus hard to differentiate. In this study, we explored the performance of a Siamese neural network using contrastive loss by learning to push apart embeddings of bumblebee species pair that are dissimilar, and pull together similar embeddings. Our experimental results show a 61% F1-score on zero-shot instances, a performance showing 11% improvement on samples of classes that share intersections with the training set.

Toward Mesh-Invariant 3D Generative Deep Learning with Geometric Measures

3D generative modeling is accelerating as the technology allowing the capture of geometric data is developing. However, the acquired data is often inconsistent, resulting in unregistered meshes or point clouds. Many generative learning algorithms require correspondence between each point when comparing the predicted shape and the target shape. We propose an architecture able to cope with different parameterizations, even during the training phase. In particular, our loss function is built upon a kernel-based metric over a representation of meshes using geometric measures such as currents and varifolds. The latter allows to implement an efficient dissimilarity measure with many desirable properties such as robustness to resampling of the mesh or point cloud. We demonstrate the efficiency and resilience of our model with a generative learning task of human faces.

Metric Nearness Made Practical

Given a square matrix with noisy dissimilarity measures between pairs of data samples, the metric nearness model computes the best approximation of the matrix from a set of valid distance metrics. Despite its wide applications in machine learning and data processing tasks, the model faces non-trivial computational requirements in seeking the solution due to the large number of metric constraints associated with the feasible region. Our work designed a practical approach in two stages to tackle the challenge and improve the model's scalability and applicability. The first stage computes a fast yet high-quality approximate solution from a set of isometrically embeddable metrics, further improved by an effective heuristic. The second stage refines the approximate solution with the Halpern-Lions-Wittmann-Bauschke projection algorithm, which converges quickly to the optimal solution. In empirical evaluations, the proposed approach runs at least an order of magnitude faster than the state-of-the-art solutions, with significantly improved scalability, complete conformity to constraints, less memory consumption, and other desirable features in real applications.

Decomposed Dissimilarity Measure for Evaluation of Digital Image Denoising

A new approach to the evaluation of digital image denoising algorithms is presented. In the proposed method, the mean absolute error (MAE) is decomposed into three components that reflect the different cases of denoising imperfections. Moreover, aim plots are described, which are designed to be a very clear and intuitive form of presentation of the new decomposed measure. Finally, examples of the application of the decomposed MAE and the aim plots in the evaluation of impulsive noise removal algorithms are presented. The decomposed MAE measure is a hybrid of the image dissimilarity measure and detection performance measures. It provides information about sources of errors such as pixel estimation errors, unnecessary altered pixels, or undetected and uncorrected distorted pixels. It measures the impact of these factors on the overall correction performance. The decomposed MAE is suitable for the evaluation of algorithms that perform a detection of the distortion that affects only a certain fraction of the image pixels.