Search

Summary Critical systems are progressively abandoning the traditional isolated and closed architectures, and adopting more federated solutions, in order to deal with orchestrated decision making within large-scale infrastructures. Such an increasing connectivity and the possibility of dynamically integrate constituents in a seamless manner by means of a decoupling middleware solution are causing the flouring of novel and previously unseen security threats, such as internal attacks conducted by camouflaged and/or compromised federated systems. Trust management is the most efficient way for dealing with such attacks, so that each constituent computes a trust degree of the other interacting ones based on the direct experiences and of collected reputation scores. An adversary may negatively affect the overall process with false reputations, which must not be considered when estimating a trust degree. Our work combines a multi-criteria linguistic fuzzy term formulation of the trust degree with the concept of entropy for measuring the divergence of certain scores from the other ones and to avoid to consider them during reputation aggregation. A set of experiments have been conducted in order to measure the quality and effectiveness of the presented approach.

Knowledge-based solution for automatic schema mapping to manage data heterogeneity.Automatic ontology extraction and semantic inference for novel Big Data analytics.Integration with publish/subscribe services for large-scale analytics infrastructures. Big Data analytics is considered an imperative aspect to be further improved in order to increase the operating margin of both public and private enterprises, and represents the next frontier for their innovation, competition, and productivity. Big Data are typically produced in different sectors of the above organizations, often geographically distributed throughout the world, and are characterized by a large size and variety. Therefore, there is a strong need for platforms handling larger and larger amounts of data in contexts characterized by complex event processing systems and multiple heterogeneous sources, dealing with the various issues related to efficiently disseminating, collecting and analyzing them in a fully distributed way.In such a scenario, this work proposes a way to overcome two fundamental issues: data heterogeneity and advanced processing capabilities. We present a knowledge-based solution for Big Data analytics, which consists in applying automatic schema mapping to face with data heterogeneity, as well as ontology extraction and semantic inference to support innovative processing. Such a solution, based on the publish/subscribe paradigm, has been evaluated within the context of a simple experimental proof-of-concept in order to determine its performance and effectiveness.

Lack of knowledge of the prior distribution in classification problems that operate on small data sets may make the application of Bayes' rule questionable. Uniform or arbitrary priors may provide classification answers that, even in simple examples, may end up contradicting our common sense about the problem. Entropic priors (EPs), via application of the maximum entropy (ME) principle, seem to provide good objective answers in practical cases leading to more conservative Bayesian inferences. EP are derived and applied to classification tasks when only the likelihood functions are available. In this paper, when inference is based only on one sample, we review the use of the EP also in comparison to priors that are obtained from maximization of the mutual information between observations and classes. This last criterion coincides with the maximization of the KL divergence between posteriors and priors that for large sample sets leads to the well-known reference (or Bernardo's) priors. Our comparison on single samples considers both approaches in prospective and clarifies differences and potentials. A combinatorial justification for EP, inspired by Wallis' combinatorial argument for entropy definition, is also included. The application of the EP to sequences (multiple samples) that may be affected by excessive domination of the class with the maximum entropy is also considered with a solution that guarantees posterior consistency. An explicit iterative algorithm is proposed for EP determination solely from knowledge of the likelihood functions. Simulations that compare EP with uniform priors on short sequences are also included.

Despite the large diffusion and use of embedding generated through Word2Vec, there are still many open questions about the reasons for its results and about its real capabilities. In particular, to our knowledge, no author seems to have analysed in detail how learning may be affected by the various choices of hyperparameters. In this work, we try to shed some light on various issues focusing on a typical dataset. It is shown that the learning rate prevents the exact mapping of the co-occurrence matrix, that Word2Vec is unable to learn syntactic relationships, and that it does not suffer from the problem of overfitting. Furthermore, through the creation of an ad-hoc network, it is also shown how it is possible to improve Word2Vec directly on the analogies, obtaining very high accuracy without damaging the pre-existing embedding. This analogy-enhanced Word2Vec may be convenient in various NLP scenarios, but it is used here as an optimal starting point to evaluate the limits of Word2Vec.

Bayesian-directed acyclic discrete-variable graphs are reduced to a simplified normal form made up of only replicator units (or equal constraint units), source, and single-input/single-output blocks. In this framework, the same adaptation algorithm can be applied to all the parametric blocks. We obtain and compare adaptation rules derived from a constrained maximum likelihood formulation and a minimum Kullback-Leibler divergence criterion using Karush-Kuhn-Tucker conditions. The learning algorithms are compared with two other updating equations based on localized decisions and on a variational approximation, respectively. The performance of the various algorithms is verified on synthetic data sets for various architectures. Factor graphs in reduced normal form provide an appealing framework for rapid deployment of Bayesian-directed graphs in the applications.

Due to their decoupling and scalability properties, multicast technologies are widely adopted in all the software projects aiming at integrating off-the-shelf systems, in order to compose large-scale complex infrastructures. This also holds within the context of mission critical systems, where strong requirements for reliable and timely data sharing are imposed. In the current literature, reliable multicast is always achieved at the expenses of violations of the temporal constraints, since retransmissions are used to recover lost messages. When timeliness requirements assume the same importance as the reliability ones, the techniques based on spatial redundancy, such as forward error correction, are preferable. However, such coding techniques are scarcely adopted within multicast services due to the difficulties related to the optimal tuning of the introduced redundancy. In this paper, we present a solution applying a proper coding scheme that jointly achieves reliability and timeliness in all those scenarios involving tree-based multicasting over the Internet. Such a solution employs game theory in order to select the best locations within the multicast tree where coding operations have to be performed. We prove the quality of this solution by using a series of simulations running on OMNET++, that compare the achievable quality with respect to the optimal solution of the underlying optimization problem.

Internet of Things (IoT) is a promising technology for improving our lives and society by integrating smart devices in our environment and paving the way for novel ICT application, spanning from smart cities to energy efficiency and home automation. However, such a vision encompasses the availability of thousands of smart devices, or even more, that continuously exchange a huge volume of data among each other and with cloud-based services, raising a big data problem. Such a problem can be approached by properly applying data fusion practices within an IoT infrastructure. Due to the characteristics and peculiarities of the communications among smart devices within the IoT, an event-based data fusion is needed, where devices exchange notifications of events among each others. Such data fusion should be focused on special devices where notification heterogeneity, and data source trust issues have to be faced with. Accordingly, the contribution of this work is proposing (i) a novel broker-less event-based communication protocol specifically tailored to sensors with constrained resources, (ii) a solution for flexible event-based communications among heterogeneous data sources, and (iii) an approach based on the theory of evidence for data fusion processes that depend on the matching and trust degree of the data to be fused.

The ever-increasing number of IoT applications and cyber–physical services is introducing significant challenges associated to their cyber-security. Due to the constrained nature of the involved devices, some heavier computational tasks, such as deep traffic inspection and classification, essential for implementing automatic attack detection systems, are moved on specialized “edge” devices, in order to distribute the processing intelligence near to the data sources. These edge devices are mainly capable of effectively running pre-built classification models but have not enough storage and processing capabilities to build and upgrade such models from huge volumes of field training data, imposing a serious barrier to the deployment of such solutions. This work leverages the flexibility of cloud-based architectures, together with the recent advancements in the area of large-scale machine learning for shifting the more computationally-expensive and storage-demanding operations to the cloud in order to benefit of edge computing capabilities only for effectively performing traffic classification based on sophisticated Extreme Learning Machines models that are pre-built over the cloud.

This paper presents a multi-camera target tracking system based on factor graphs, designed to estimate the state of moving targets using a message propagation framework. Inexpensive cameras deployed in the area of interest are the main sensory modality. We discuss the transformation geometry between world points and camera pixels and how the uncertainties in the homographies affect the estimation. The framework has been tested on real data coming from a harbor to track moving vessels.

Detection algorithms for electroencephalography (EEG) data typically employ handcrafted features that take advantage of the signal's specific properties. In the field of interictal epileptic discharge (IED) detection, the feature representation that provides optimal classification performance is still an unresolved issue. In this paper, we consider deep learning for automatic feature generation from epileptic intracranial EEG data in the time domain. Specifically, we consider convolutional neural networks (CNNs) in a subject independent fashion and demonstrate that meaningful features, representing IEDs are automatically learned. The resulting model achieves state of the art classification performance, provides insights for the different types of IEDs within the group, and is invariant to time differences between the IEDs. This study suggests that automatic feature generation via deep learning is suitable for IEDs and EEG in general.

Gesture recognition is a hot topic in research, due to its appealing applications in real-life contexts, from remote control to assistive robotics. In this paper we focus on grasping gestures recognition. This kind of gestures is particularly interesting, because it requires not only analyzing hand trajectories, but also fingers position and fingertip forces, of utmost importance in manipulation tasks. We used a discrete HMM-based model for gesture recognition. Input codebooks for the model are gesture elementary phases, obtained through a LLS-regression segmentation algorithm, and feature vectors representing hand position over time.

We present a novel appearance-based approach for pose estimation of a human hand using the point clouds provided by the low-cost Microsoft Kinect sensor. Both the free-hand case, in which the hand is isolated from the surrounding environment, and the hand-object case, in which the different types of interactions are classified, have been considered. The pose estimation is obtained by applying a modified version of the Iterative Closest Point (ICP) algorithm to the synthetic models. The proposed framework uses a “pure” point cloud as provided by the Kinect sensor without any other information such as RGB values or normal vector components.

Bayesian clustering implemented on a small Factor Graph is utilized in this work to perform associative recall and pattern recognition on images. The network is trained using a maximum likelihood algorithm on images from a standard data set. The two-class labels are fused with the image data into a unique hidden variable. Performances are evaluated in terms of Kullback-Leibler (KL) divergence between forward and backward messages for images and labels. These experiments reveal the nature of the representation that the learning algorithm builds in the hidden variable.

Internet of Things is a futuristic vision of networked objects able to take smart decisions and to assist humans in their daily activities. The founding factor of such a vision is the capacity of objects to exchange data by means of wireless networks, and the applied communication pattern is event-driven and multicast. Publish/subscribe services are a promising communication technology for building solutions according to the Internet of Things' vision. The available solutions consist in the adaptation of widely-known products for publish/subscribe services taken from other contexts and application domains, without considering the key peculiarities of the Internet of Things. Our driving idea is to consider beaconing as the basic communication means, and to build a publish/subscribe service based on it. Our solution can be used stand-alone for event-driven communications and/or integrated within standardized protocols to provide basic communication capabilities. We present a preliminary set of experiments showing the efficiency of our solution. Refereed/Peer-reviewed

Cloud platforms encompass a large number of storage services that can be used to manage the needs of customers. Each of these services, offered by a different provider, is characterized by specific features, limitations and prices. In presence of multiple options, it is crucial to select the best solution fitting the requirements of the customer in terms of quality of service and costs. Most of the available approaches are not able to handle uncertainty in the expression of subjective preferences from customers, and can result in wrong (or sub-optimal) service selections in presence of rational/selfish providers, exposing untrustworthy indications concerning the quality of service levels and prices associated to their offers. In addition, due to its multi-objective nature, the optimal service selection process results in a very complex task to be managed, when possible, in a distributed way, for well-known scalability reasons. In this work, we aim at facing the above challenges by proposing three novel contributions. The fuzzy sets theory is used to express vagueness in the subjective preferences of the customers. The service selection is resolved with the distributed application of fuzzy inference or Dempster-Shafer theory of evidence. The selection strategy is also complemented by the adoption of a game theoretic approach for promoting truth-telling ones among service providers. We present empirical evidence of the proposed solution effectiveness through properly crafted simulation experiments.

When given a single frame of the video, humans can not only interpret the content of the scene, but also they are able to forecast the near future. This ability is mostly driven by their rich prior knowledge about the visual world, both in terms of (i) the dynamics of moving agents, as well as (ii) the semantic of the scene. In this work we exploit the interplay between these two key elements to predict scene-specific motion patterns. First, we extract patch descriptors encoding the probability of moving to the adjacent patches, and the probability of being in that particular patch or changing behavior. Then, we introduce a Dynamic Bayesian Network which exploits this scene specific knowledge for trajectory prediction. Experimental results demonstrate that our method is able to accurately predict trajectories and transfer predictions to a novel scene characterized by similar elements., Accepted to ECCV 2016

Matching cross-view images is challenging because the appearance and viewpoints are significantly different. While low-level features based on gradient orientations or filter responses can drastically vary with such changes in viewpoint, semantic information of images however shows an invariant characteristic in this respect. Consequently, semantically labeled regions can be used for performing cross-view matching. In this paper, we therefore explore this idea and propose an automatic method for detecting and representing the semantic information of an RGB image with the goal of performing cross-view matching with a (non-RGB) geographic information system (GIS). A segmented image forms the input to our system with segments assigned to semantic concepts such as traffic signs, lakes, roads, foliage, etc. We design a descriptor to robustly capture both, the presence of semantic concepts and the spatial layout of those segments. Pairwise distances between the descriptors extracted from the GIS map and the query image are then used to generate a shortlist of the most promising locations with similar semantic concepts in a consistent spatial layout. An experimental evaluation with challenging query images and a large urban area shows promising results.

We build a multi-layer architecture using the Bayesian framework of the Factor Graphs in Reduced Normal Form (FGrn). This model allows great modularity and unique localized learning equations. The multi-layer architecture implements a hierarchical data representation that via belief propagation can be used for learning and inference in pattern completion, correction and classification. We apply the framework to images extracted from a standard data set.

System modeling with Probabilistic Graphical Models (PGM) has become increasingly popular in the last years. In this paper we design a Multiple Target Tracker based on the probabilistic architecture of Normal Factor Graph. Belief propagation makes best use of data coming from different branches of the graph and yields the tracks via messages fusion. The issues of data association, track life-cycle management and data fusion from heterogeneous sensor modalities are resolved at each time step by propagating and combining forward and backward probabilistic messages. Inexpensive cameras deployed in the scene under surveillance are the primary sensor modality, even if the framework has been designed to receive data from a wide range of sensors such as Radars, Infrared cameras, etc. The framework has been tested by calculating the tracks of different ships moving in an harbour framed by three cameras. © 2014 IEEE.

Propagation of Gaussian belief messages in factor graphs in normal form is applied to data fusion for tracking moving objects in maritime scenarios, as crowded harbors. The data are yielded by multiple cameras, deployed in the region under surveillance, and AIS system, wherever is available. The track model and the estimates coming from the sensors are integrated bi-directionally, providing a flexible framework for comprehensive inference. The framework is applied to tracking a large cargo ship in a harbor from frames recorded with three commercial cameras. © 2014 IEEE.

In recent years there has been an increasing demand of systems that automatically manage and control the state of large critical areas, such as airports, harbors, parking lots, etc. The framework of the Bayesian Factor Graphs to target fusion seems to be quite promising with respect to classical approaches because of its modularity and because it can naturally integrate very heterogeneous sources of information. The system presented in this paper fuses real-time data coming from various sensors, along with estimates coming from the tracked object models (if available). All the information is merged within environmental constraints in order to provide the best estimate of the state of a moving object. Factor graphs allow the information to flow bidirectionally, to predict the future, or to strengthen our knowledge of the past. In this paper we focus on camera sensors, deployed along the area of interest. The information is merged into the factor graph after geometric inversion and covariance estimate. The problem of automatic localization of moving objects on the images is also addressed. The framework has been tested on a parking area, where states are estimated, accuracy is assessed and considerations about the framework are provided. © 2014 IEEE.

Interaction analysis of ships mooring and maneuvering in harbors is pursued in this paper by using Bayesian probabilistic models. A number of ship-to-ship interactions are deduced from the navigation rules in port areas, and then used to train different Event-based Dynamic Bayesian Networks (E-DBNs). When data of two interacting ships are injected into the network, inference is performed in order to verify if the interaction between the vessels is known or not, and in the latter case actions to preserve the port safety can be taken. Results are drawn in the final part of the paper by including into the networks data provided by a simulator of realistic trajectories relative to an existing port.

We study the behavior of anti-Hebbian synapses at a neural node that contains the standard sigmoidal nonlinearity. The criterion generalizes the idea already discussed by one of the authors for linear networks and consists in removing the correlation between the input to the synapse, and the node output. We show how the solution, just as for the linear case, is unique and can be learned with a standard anti-Hebbian rule. We suggest how these synapses can be embedded in fully self-organizing networks to generate orthogonal nonlinear components and be used for multidimensional approximation.

We focus on learning the probability matrix for discrete random variables in factor graphs. We review the problem and its variational approximation and, via entropic priors, we show that soft quantization can be included in a probabilistically-consistent fashion in a factor graph that learns the mutual relationship among the variables involved. The framework is explained with reference the ”Tipper” example and the results of a Matlab simulation are included.

In pattern classification problems lack of knowledge about the prior distribution is typically filled up with uniform priors. However this choice may lead to unsatisfactory inference results when the amount of observed data is scarce. The application of Maximum Entropy (ME) principle to prior determination results in the so-called en-tropic priors, which provide a much more cautious inference in comparison to uniform priors. The idea, introduced mainly within the context of theoretical physics, is applied here to signal processing scenarios. We derive efficient formulas for computing and updating entropic priors when the the likelihoods follow on Independent, Markov and Hidden Markov models and we apply them to a target-track classification task.

The paper shows how to build an associative memory from a finite list of ex-amples. By means of a fully-blown example, it is demostrate how a probabilistic Bayesian factor graph can integrate naturally the discrete information contained in the list with smooth inference.

This paper proposes a source-traffic based model to estimate jointly packet losses and delays statistical behavior of a network path. The approach relies on a hidden Markov model built on real-traffic information. The effectiveness of the model is evaluated over different real heterogeneous network scenarios. Our experimental results show that the model captures average (long-term) and conditional (short-term) statistics that in most cases are typical of the single scenario. Preliminary results about prediction on a sample path as well as investigation on the use of the same model across different scenarios are given.

One of the most effective ways of modeling vector data for unsupervised pattern classification or coding, is to assume that the observations are the result of picking randomly out of a fixed set of different distributions. In this paper we propose to perform the unsupervised estimation of the mixture density underlying the data as the problem of separating multiclass sources. Assuming in each class independent components, standard linear independent component analysis (ICA) can be adopted in the recently extended mode which provides signal reconstruction for a multiclass mixture. Unfortunately, in practical problems the class densities necessary to match the experimental distributions must be degenerate or poorly conditioned. In this paper we approach the problem by assuming from the beginning sources which have either rank-deficient distributions or show very concentrated eigenvalues. The class membership of each point is based on a distance measure from the hyperplanes and on the likelihood on each hyperplane. The independent components are then searched within each subspace. We present results of the algorithm on synthetic distributions with various degrees of degeneracy. Our results are promising for feature extraction applications.

The support vector machine (SVM) constitutes one of the most powerful methods for constructing a mathematical model on the basis of a given number of training examples. SVM training requires that we solve a quadratic optimization problem; this step is usually performed by means of existing software packages. Such a black-box approach may be undesirable. In this paper we introduce a simple iterative algorithm for SVM training which compares well with some typical software packages, can be simply implemented, and has minimal memory requirements. It addresses the problem of regression estimation and utilizes ideas similar to those proposed by J. Platt (1998) for training binary SVM.

The Support Vector Machine, introduced in [1] as a practical implementation of the principle of structural risk minimization, constitutes one of the most promising methods for constructing a mathematical model only on the base of a limited amount of measured data. In this paper, we consider the application of this method to the problem of nonparametric binary hypothesis testing (bayesian setting); the main contribution of this paper is the derivation of the Support Vector algorithm in the case of a generic convex approximation of the binary risk function.

Most support vector (SV) methods proposed in the recent literature can be viewed in a unified framework with great flexibility in terms of the choice of the kernel functions and their constraints. We show that all these problems can be solved within a unique approach if we are equipped with a robust method for finding a sparse solution of a linear system. Moreover, for such a purpose, we propose an iterative algorithm that can be simply implemented. Finally, we compare the classical SV approach with other, recently proposed, cross-correlation based, alternative methods. The simplicity of their implementation and the possibility of exactly calculating their computational complexity constitute important advantages in a real-time signal processing scenario.

The semiparametric Support Vector Machine (SVM) has recently been introduced as a generalization of the classical SVM to the case in which some a priori knowledge about the considered problem is available. Semiparametric SVM training requires that we solve an optimization problem very similar (it only imposes a larger number of equality constraints) to that to be solved for classical SVM training. In both cases training is usually performed by means of existing software packages. Since this black-box approach may be undesirable, with reference to the classical SVM, some simple and explicit algorithms, difficult to extend to the semiparametric case, have recently been proposed. In this paper we introduce a simple iterative algorithm for semiparametric SVM training which compares well with some typical software packages, can be simply implemented and has minimal memory requirements.

A binary factorial code is a transformation that maps the input space into a binary string with independent bits. We apply independent component analysis, formulated as a density search, to the pursuit of such a code by considering a set of desired output bi-modal density functions. The idea is tested on a real image with the coder used also in a reversed order as a generative model. The results are also compared to a principal component analyzer. Fantasy blocks of the image transforms provide interesting clues on the nature of the learned representations.

Performance of real-time applications on network communication channels are strongly related to losses and temporal delays. Several studies showed that these network features may be correlated and present a certain degree of memory such as bursty losses and delays. The memory and the statistical dependence between losses and temporal delays suggest that the channel may be well modelled by a Dynamic Bayesian Network with an appropriate hidden variable that captures the current state of the network. In this paper we propose a Bayesian model that, trained with a version of the EM-algorithm, seems to be effective in modelling typical channel behaviors.

Performance of real-time applications on network communication channels are strongly related to losses and temporal delays. Several studies showed that these network features may be correlated and exhibit a certain degree of memory such as bursty losses and delays. The memory and the statistical dependence between losses and temporal delays suggest that the channel may be well modelled by a Hidden Markov Model (HMM) with appropriate hidden variables that capture the current state of the network. In this paper we discuss on the effectiveness of using an HMM to model jointly loss and delay behavior of real communication channel. Excellent performance in modelling typical channel behavior in a set of real packet links are observed. The system parameters are found via a modified version of the EM algorithm. Hidden state analysis shows how the state variables characterize channel dynamics. State-sequence estimation is obtained by use of the Viterbi algorithm. Real-time modelling of the channel is the first step to implement adaptive communication strategies.

Human ability to foresee the near future plays a key role in everyone's life to prevent potentially dangerous situations. To be able to make predictions is crucial when people have to interact with the surrounding environment. Modeling such capability can lead to the design of automated warning systems and provide moving robots with an intelligent way of interaction with changing situation. In this work we focus on a typical urban human-scene where we aim at predicting an agent's behavior using a stochastic model. In this approach, we fuse the various factors that would contribute to a human motion in different contexts. Our method uses previously observed trajectories to build point-wise circular distributions that after combination, provide a statistical smooth prediction towards the most likely areas. More specifically, a ray-launching procedure, based on a semantic segmentation, gives a coarse scene representation for collision avoidance; a nearly-constant velocity dynamic model smooths the acceleration progression and knowledge of the agent's destination may further steer the path prediction. Experimental results in structured scenes, validate the effectiveness of the method in predicting paths in comparison to actual trajectories.