Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9627063
A. Jousselme, T. Wickramarathne, P. Kowalski
In decision-making under uncertainty, the objective of measuring information value is to assist a decision-maker toward making good decisions by generating necessary metadata about the information that is being considered for the decision-making task. Epistemic decisions about which sources to trust or query are critical for a decision-maker when the end-goal (or final) decisions are to be made using limited number of information sources. This is even more critical in a multi-intelligence context, where information sources are highly heterogeneous, prone to errors, partially informed, deceptive, or even malicious. In these contexts, the ability to distinguish between aleatory and epistemic uncertainty (ignorance) becomes a fundamental requirement. In this paper, we propose some extensions to classical measures of value of information for imprecise belief states represented by belief functions relying on a general observation model. The proposed measures allow a decision-maker to highlight critical parameters, such as the probability of source reliability and the degree of confidence expressed by the source. We compare several decision models and illustrate the use of proposed measures in a maritime surveillance scenario, where the decision-maker has to make a rational selection of information sources that consists of both physical sensors and human sources. We conclude by providing some insights on future research directions to expand this preliminary exploration.
{"title":"Toward Measuring Information Value in a Multi-Intelligence Context","authors":"A. Jousselme, T. Wickramarathne, P. Kowalski","doi":"10.23919/fusion49465.2021.9627063","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9627063","url":null,"abstract":"In decision-making under uncertainty, the objective of measuring information value is to assist a decision-maker toward making good decisions by generating necessary metadata about the information that is being considered for the decision-making task. Epistemic decisions about which sources to trust or query are critical for a decision-maker when the end-goal (or final) decisions are to be made using limited number of information sources. This is even more critical in a multi-intelligence context, where information sources are highly heterogeneous, prone to errors, partially informed, deceptive, or even malicious. In these contexts, the ability to distinguish between aleatory and epistemic uncertainty (ignorance) becomes a fundamental requirement. In this paper, we propose some extensions to classical measures of value of information for imprecise belief states represented by belief functions relying on a general observation model. The proposed measures allow a decision-maker to highlight critical parameters, such as the probability of source reliability and the degree of confidence expressed by the source. We compare several decision models and illustrate the use of proposed measures in a maritime surveillance scenario, where the decision-maker has to make a rational selection of information sources that consists of both physical sensors and human sources. We conclude by providing some insights on future research directions to expand this preliminary exploration.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123255906","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9626848
Wei Li, Deqiang Han, Xiaojing Fan, Bo Dong
The determination of the basic belief assignment (BBA) is an important yet difficult problem in evidence theory. In this paper, some BBA determination methods using density estimation that can directly generate compound focal elements for ambiguous classes are proposed, including the Gaussian Mixture Model (GMM) based and Generative Adversarial Network (GAN) based methods. Experimental results of evidence combination based pattern classification on various UCI data sets show that our new proposed methods are rational and can effectively improve the accuracy of fusion based pattern classification.
{"title":"Method of Basic Belief Assignment Determination Based on Density Estimation of Ambiguous Samples","authors":"Wei Li, Deqiang Han, Xiaojing Fan, Bo Dong","doi":"10.23919/fusion49465.2021.9626848","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9626848","url":null,"abstract":"The determination of the basic belief assignment (BBA) is an important yet difficult problem in evidence theory. In this paper, some BBA determination methods using density estimation that can directly generate compound focal elements for ambiguous classes are proposed, including the Gaussian Mixture Model (GMM) based and Generative Adversarial Network (GAN) based methods. Experimental results of evidence combination based pattern classification on various UCI data sets show that our new proposed methods are rational and can effectively improve the accuracy of fusion based pattern classification.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122777107","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9627056
Audrey Cuillery, F. Gland
In multitarget tracking, many particle approximations are available to sample from the filtering density, with the effect that multitarget particles are obtained by discarding or replicating globally the existing multitarget particles, i.e. the particles for all the different targets are replicated from the same multitarget particle. A better design would be to produce shuffled multitarget particles such that the particle for each different target can be replicated from a different multitarget particle. An efficient solution has been proposed by Ubéda–Medina et al. under a posterior independence assumption that is almost never met in practical situations. The objective of this work is to propose another solution that does not rely on the posterior independence assumption. This new solution is based on introducing an auxiliary Markov transition, and is seen as an extension of the auxiliary particle filter.
{"title":"Particle filters with auxiliary Markov transition. Application to crossover and to multitarget tracking","authors":"Audrey Cuillery, F. Gland","doi":"10.23919/fusion49465.2021.9627056","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9627056","url":null,"abstract":"In multitarget tracking, many particle approximations are available to sample from the filtering density, with the effect that multitarget particles are obtained by discarding or replicating globally the existing multitarget particles, i.e. the particles for all the different targets are replicated from the same multitarget particle. A better design would be to produce shuffled multitarget particles such that the particle for each different target can be replicated from a different multitarget particle. An efficient solution has been proposed by Ubéda–Medina et al. under a posterior independence assumption that is almost never met in practical situations. The objective of this work is to propose another solution that does not rely on the posterior independence assumption. This new solution is based on introducing an auxiliary Markov transition, and is seen as an extension of the auxiliary particle filter.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128967658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9626974
Magnus Malmström, I. Skog, Daniel Axehill, F. Gustafsson
Despite the great success of neural networks (NN) in many application areas, it is still not obvious how to integrate an NN in a sensor fusion framework. The reason is that the computation of the for fusion required variance of NN is still a rather immature area. Here, we apply a methodology from system identification where uncertainty of the parameters in the NN are first estimated in the training phase, and then this uncertainty is propagated to the output in the prediction phase. This local approach is based on linearization, and it implicitly assumes a good signal-to-noise ratio and persistency of excitation. We illustrate the proposed method on a fundamental problem in advanced driver assistance systems (ADAS), namely to estimate the tire-road friction. This is a single input single output static nonlinear relation that is simple enough to provide insight and it enables comparisons with other parametric approaches. We compare both to existing methods for how to assess uncertainty in NN and standard methods for this problem, and evaluate on real data. The goal is not to improve on simpler methods for this particular application, but rather to validate that our method is on par with simpler model structures, where output variance is immediately provided.
{"title":"Modeling of the tire-road friction using neural networks including quantification of the prediction uncertainty","authors":"Magnus Malmström, I. Skog, Daniel Axehill, F. Gustafsson","doi":"10.23919/fusion49465.2021.9626974","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9626974","url":null,"abstract":"Despite the great success of neural networks (NN) in many application areas, it is still not obvious how to integrate an NN in a sensor fusion framework. The reason is that the computation of the for fusion required variance of NN is still a rather immature area. Here, we apply a methodology from system identification where uncertainty of the parameters in the NN are first estimated in the training phase, and then this uncertainty is propagated to the output in the prediction phase. This local approach is based on linearization, and it implicitly assumes a good signal-to-noise ratio and persistency of excitation. We illustrate the proposed method on a fundamental problem in advanced driver assistance systems (ADAS), namely to estimate the tire-road friction. This is a single input single output static nonlinear relation that is simple enough to provide insight and it enables comparisons with other parametric approaches. We compare both to existing methods for how to assess uncertainty in NN and standard methods for this problem, and evaluate on real data. The goal is not to improve on simpler methods for this particular application, but rather to validate that our method is on par with simpler model structures, where output variance is immediately provided.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124750265","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9627019
Desheng Zhang, Wujun Li, Wei Yi
This paper addresses the joint tracking and shape estimation (JTSE) problem of elliptical extended targets in low signal-to-noise (SNR) scenarios using multi-frame joint processing. Considering the weak target echoes and unknown parameters of elliptical extended targets, it is challenging to achieve effective detection and tracking. To solve these problems, a multi-frame tracking and shape estimation (MF-JTSE) method is proposed. This method achieves accurate estimation of motion trajectories and shape parameters including semi-axis lengths simultaneously for unknown priori information. By comparing with single-frame joint tracking and shape estimation (SF-JTSE) methods, simulation results show that the proposed algorithm is able to achieve superior tracking performance and estimation accuracy for extended targets in low SNR scenarios.
{"title":"Multi-Frame Joint Tracking and Shape Estimation Method for Weak Extended Targets","authors":"Desheng Zhang, Wujun Li, Wei Yi","doi":"10.23919/fusion49465.2021.9627019","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9627019","url":null,"abstract":"This paper addresses the joint tracking and shape estimation (JTSE) problem of elliptical extended targets in low signal-to-noise (SNR) scenarios using multi-frame joint processing. Considering the weak target echoes and unknown parameters of elliptical extended targets, it is challenging to achieve effective detection and tracking. To solve these problems, a multi-frame tracking and shape estimation (MF-JTSE) method is proposed. This method achieves accurate estimation of motion trajectories and shape parameters including semi-axis lengths simultaneously for unknown priori information. By comparing with single-frame joint tracking and shape estimation (SF-JTSE) methods, simulation results show that the proposed algorithm is able to achieve superior tracking performance and estimation accuracy for extended targets in low SNR scenarios.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121936719","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9627007
J. D. Villiers, G. Pavlin, J. Ziegler, A. Jousselme, P. Costa, Eric Blasch, Kathryn B. Laskey, C. Laudy, A. D. Waal, J.-H. Cho
To be employed effectively, an information fusion system must be trusted. As information fusion systems grow more complex, the question of trust grows more pressing. This paper addresses the question of evaluating trust in the uncertainty representation and reasoning aspects of an information fusion system. The four main aspects of trust in information fusion systems considered in this paper are: (1) how trust manifests itself in an information fusion system; (2) the temporal aspects of trust and its effect on the decision process; (3) the uncertainty associated with trust; and (4) exploring the evaluation of the uncertainty associated with trust using the Uncertainty Representation and Reasoning Framework (URREF) ontology and other trust related ontologies. The focus of the paper is on measuring trust related uncertainty to engage users towards adopting information fusion systems in mission systems. Mapping trust constructs into the Uncertainty Representation and Reasoning Framework provides measurable criteria for trust uncertainty analysis and evaluation. The ideas put forward in this paper serve as a foundation for further discussions within the Evaluation Techniques for Uncertainty Representation Working Group (ETURWG) on how to evaluate the uncertainty aspects of trust in information fusion systems, and to further cement application of the URREF ontology for the evaluation of trust uncertainty.
为了使信息融合系统得到有效利用，信息融合系统必须是可信的。随着信息融合系统变得越来越复杂，信任问题也变得越来越紧迫。本文研究了信息融合系统在不确定性表示和推理方面的信任评估问题。本文研究了信息融合系统中信任的四个主要方面:(1)信任如何在信息融合系统中表现出来;(2)信任的时间层面及其对决策过程的影响;(3)与信任相关的不确定性;(4)利用不确定性表示与推理框架(URREF)本体和其他与信任相关的本体探索与信任相关的不确定性的评估。本文的重点是测量与信任相关的不确定性，以吸引用户在任务系统中采用信息融合系统。将信任结构映射到不确定性表示和推理框架中，为信任不确定性分析和评估提供了可测量的标准。本文提出的思想为不确定性表示评估技术工作组(Evaluation Techniques for Uncertainty Representation Working Group, ETURWG)进一步讨论如何评估信息融合系统中信任的不确定性方面奠定了基础，并进一步巩固了URREF本体在评估信任不确定性方面的应用。
{"title":"Uncertainty Evaluation of Temporal Trust in a Fusion System Using the URREF Ontology","authors":"J. D. Villiers, G. Pavlin, J. Ziegler, A. Jousselme, P. Costa, Eric Blasch, Kathryn B. Laskey, C. Laudy, A. D. Waal, J.-H. Cho","doi":"10.23919/fusion49465.2021.9627007","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9627007","url":null,"abstract":"To be employed effectively, an information fusion system must be trusted. As information fusion systems grow more complex, the question of trust grows more pressing. This paper addresses the question of evaluating trust in the uncertainty representation and reasoning aspects of an information fusion system. The four main aspects of trust in information fusion systems considered in this paper are: (1) how trust manifests itself in an information fusion system; (2) the temporal aspects of trust and its effect on the decision process; (3) the uncertainty associated with trust; and (4) exploring the evaluation of the uncertainty associated with trust using the Uncertainty Representation and Reasoning Framework (URREF) ontology and other trust related ontologies. The focus of the paper is on measuring trust related uncertainty to engage users towards adopting information fusion systems in mission systems. Mapping trust constructs into the Uncertainty Representation and Reasoning Framework provides measurable criteria for trust uncertainty analysis and evaluation. The ideas put forward in this paper serve as a foundation for further discussions within the Evaluation Techniques for Uncertainty Representation Working Group (ETURWG) on how to evaluate the uncertainty aspects of trust in information fusion systems, and to further cement application of the URREF ontology for the evaluation of trust uncertainty.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131364006","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9626858
Aranee Balachandran, R. Tharmarasa
This paper analyzes the problem of incorporating multipath measurements from an unknown reflection surface for improving the tracking result of a single target. The characteristic of the reflection surface, such as location and the slope, should be known in order to use the multipath measurements for track initialization or filtering. However, in a real-world problem, the reflection surface is mostly unknown or partial information about the reflection surface is known. If the problem of estimating the unknown parameters of the reflection surface is observable with the direct and multipath measurements, then the multipath measurements from the unknown reflection surface could be used to improve tracking performance. In this paper, a tracking framework is proposed to track a single target with an unknown reflection surface, and the Fisher Information Matrix (FIM) is derived for the considered problem to examine the observability. In addition, simulation results showing the performance of multipath-assisted tracking and the performance bounds are also provided.
{"title":"Observability Analysis of Multipath Assisted Target Tracking with Unknown Reflection Surface","authors":"Aranee Balachandran, R. Tharmarasa","doi":"10.23919/fusion49465.2021.9626858","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9626858","url":null,"abstract":"This paper analyzes the problem of incorporating multipath measurements from an unknown reflection surface for improving the tracking result of a single target. The characteristic of the reflection surface, such as location and the slope, should be known in order to use the multipath measurements for track initialization or filtering. However, in a real-world problem, the reflection surface is mostly unknown or partial information about the reflection surface is known. If the problem of estimating the unknown parameters of the reflection surface is observable with the direct and multipath measurements, then the multipath measurements from the unknown reflection surface could be used to improve tracking performance. In this paper, a tracking framework is proposed to track a single target with an unknown reflection surface, and the Fisher Information Matrix (FIM) is derived for the considered problem to examine the observability. In addition, simulation results showing the performance of multipath-assisted tracking and the performance bounds are also provided.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131530257","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9626963
Sixian Zhang, Yi Yang, Meng Zhang, Pengbo Mi
The discriminative correlation filter (DCF) does not work well in complex tracking scenarios. In order to improve the accuracy of object tracking, a new correlation filter tracker is proposed. We use the different hash algorithm to screen candidate samples, reduce the number of negative samples and improve the speed and accuracy of object tracking; combine the HOG feature with color histogram feature to acquire a robust object appearance model; design an adaptive fusion function to fuse the two features to obtain a more discriminative feature and improve the discriminability of the filter. Experiments on OTB2015 show that the proposed tracker has good accuracy in complex tracking scenes such as fast motion, background clutter, illumination variation, scale variation, etc.
{"title":"A Mutli-feature Correlation Filter Tracker with Different Hash Algorithm","authors":"Sixian Zhang, Yi Yang, Meng Zhang, Pengbo Mi","doi":"10.23919/fusion49465.2021.9626963","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9626963","url":null,"abstract":"The discriminative correlation filter (DCF) does not work well in complex tracking scenarios. In order to improve the accuracy of object tracking, a new correlation filter tracker is proposed. We use the different hash algorithm to screen candidate samples, reduce the number of negative samples and improve the speed and accuracy of object tracking; combine the HOG feature with color histogram feature to acquire a robust object appearance model; design an adaptive fusion function to fuse the two features to obtain a more discriminative feature and improve the discriminability of the filter. Experiments on OTB2015 show that the proposed tracker has good accuracy in complex tracking scenes such as fast motion, background clutter, illumination variation, scale variation, etc.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132811173","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9627011
A. D'Ortenzio, C. Manes
Gaussian Mixtures (GMs) are a powerful tool for approximating probability distributions across a variety of fields. In some applications the number of GM components rapidly grows with time, so that reduction algorithms are necessary. Given a GM with a large number of components, the problem of Gaussian Mixture Reduction (GMR) consists in finding a GM with considerably less components that is not too dissimilar from the original one. There are many issues that make non trivial this problem. First of all, many dissimilarity measures exist for GMs, although most of them lack closed forms, and their numerical computation is a demanding task, especially for distributions in high dimensions. Moreover, some basic reduction actions can be simple or complex tasks depending on which dissimilarity measure is chosen. It follows that most reduction procedures proposed in the literature are made of steps that are aimed at maintaining low dissimilarity according to different measures, thus leading to a pipeline of actions that are not mutually consistent. In this paper Composite Transportation Dissimilarities are discussed and exploited to formulate a GMR framework that preserves consistency with a unique dissimilarity measure, and provides a generalization of the celebrated Runnalls’ GMR approach.
{"title":"Composite Transportation Dissimilarity in Consistent Gaussian Mixture Reduction","authors":"A. D'Ortenzio, C. Manes","doi":"10.23919/fusion49465.2021.9627011","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9627011","url":null,"abstract":"Gaussian Mixtures (GMs) are a powerful tool for approximating probability distributions across a variety of fields. In some applications the number of GM components rapidly grows with time, so that reduction algorithms are necessary. Given a GM with a large number of components, the problem of Gaussian Mixture Reduction (GMR) consists in finding a GM with considerably less components that is not too dissimilar from the original one. There are many issues that make non trivial this problem. First of all, many dissimilarity measures exist for GMs, although most of them lack closed forms, and their numerical computation is a demanding task, especially for distributions in high dimensions. Moreover, some basic reduction actions can be simple or complex tasks depending on which dissimilarity measure is chosen. It follows that most reduction procedures proposed in the literature are made of steps that are aimed at maintaining low dissimilarity according to different measures, thus leading to a pipeline of actions that are not mutually consistent. In this paper Composite Transportation Dissimilarities are discussed and exploited to formulate a GMR framework that preserves consistency with a unique dissimilarity measure, and provides a generalization of the celebrated Runnalls’ GMR approach.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133671392","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.23919/fusion49465.2021.9626978
A. D'Ortenzio, C. Manes
In many practical contexts, Gaussian Mixtures are used as density approximators due to their versatility and representation capabilities. In some scenarios, it might be convenient to approximate a set of Gaussian densities with a single one, according to criteria which aim to preserve information while reducing the model complexity. This task can be seen as a particular case of the Gaussian Mixture Reduction problem, where the goal is to find a mixture of reduced size yielding the least dissimilarity from the original mixture. From a different perspective, this can be interpreted as a data fusion process, where several Gaussian densities are fused into one. In this work, an information-theoretic class of measures will be explored in the analytical and numerical properties in order to provide insights on their nature when adopted in a Gaussian mixture reduction or data fusion process.
{"title":"Likeness-Based dissimilarity measures for Gaussian Mixture Reduction and Data Fusion","authors":"A. D'Ortenzio, C. Manes","doi":"10.23919/fusion49465.2021.9626978","DOIUrl":"https://doi.org/10.23919/fusion49465.2021.9626978","url":null,"abstract":"In many practical contexts, Gaussian Mixtures are used as density approximators due to their versatility and representation capabilities. In some scenarios, it might be convenient to approximate a set of Gaussian densities with a single one, according to criteria which aim to preserve information while reducing the model complexity. This task can be seen as a particular case of the Gaussian Mixture Reduction problem, where the goal is to find a mixture of reduced size yielding the least dissimilarity from the original mixture. From a different perspective, this can be interpreted as a data fusion process, where several Gaussian densities are fused into one. In this work, an information-theoretic class of measures will be explored in the analytical and numerical properties in order to provide insights on their nature when adopted in a Gaussian mixture reduction or data fusion process.","PeriodicalId":226850,"journal":{"name":"2021 IEEE 24th International Conference on Information Fusion (FUSION)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132444399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：