Kernel-based algorithms have been a topic of considerable interest in the machine learning community over last ten years. Their attractiveness resides their elegant treatment nonlinear problems. They successfully applied to pattern recognition, regression and density estimation. A common characteristic kernel-based methods is that they deal with kernel expansions whose number terms equals input data, making them unsuitable for online applications. Recently, several solutions proposed...

When considering the problem of unmixing hyperspectral images, most literature in geoscience and image processing areas relies on widely used linear mixing model (LMM). However, LMM may be not valid other nonlinear models need to considered, for instance, when there are multi-scattering effects or intimate interactions. Consequently, over last few years, several significant contributions have been proposed overcome limitations inherent LMM. In this paper, we present an overview recent...

Adaptive networks are suitable for decentralized inference tasks. Recent works have intensively studied distributed optimization problems in the case where nodes to estimate a single optimum parameter vector collaboratively. However, there many important applications that multitask-oriented sense multiple vectors be inferred simultaneously, collaborative manner, over area covered by network. In this paper, we employ diffusion strategies develop algorithms address multitask minimizing an...

The diffusion LMS algorithm has been extensively studied in recent years. This efficient strategy allows to address distributed optimization problems over networks the case where nodes have collaboratively estimate a single parameter vector. Problems of this type are referred as single-task problems. Nevertheless, there several practice that multitask-oriented sense optimum vector may not be same for every node. brings up issue studying performance when it is run, either intentionally or...

Spectral unmixing is central when analyzing hyperspectral data. To accomplish this task, physics-based methods have become popular because, with their explicit mixing models, they can provide a clear interpretation. Nevertheless, because of limited modeling capabilities, especially real scenes unknown complex physical properties, these may not be accurate. On the other hand, data-driven using deep learning in particular developed rapidly recent years, thanks to superior capability nonlinear...

Spectral unmixing is an important issue to analyze remotely sensed hyperspectral data. Although the linear mixture model has obvious practical advantages, there are many situations in which it may not be appropriate and could advantageously replaced by a nonlinear one. In this paper, we formulate new kernel-based paradigm that relies on assumption mixing mechanism can described of endmember spectra, with additive fluctuations defined reproducing kernel Hilbert space. This family models clear...

An operational framework is developed for testing stationarity relatively to an observation scale, in both stochastic and deterministic contexts. The proposed method based on a comparison between global local time-frequency features. originality make use of family stationary surrogates defining the null hypothesis base them two different statistical tests. first one makes suitably chosen distances spectra, whereas second implemented as one-class classifier, time- frequency features extracted...

Sparse hyperspectral unmixing from large spectral libraries has been considered to circumvent limitations of endmember extraction algorithms in many applications. This strategy often leads ill-posed inverse problems, which can benefit spatial regularization strategies. While existing methods improve the problem conditioning and promote piecewise smooth solutions, they lead nonsmooth optimization problems. Thus, efficiently introducing context remains a challenge, necessity for real world In...

The problem of learning simultaneously several related tasks has received considerable attention in domains, especially machine with the so-called multitask or to learn [1], [2]. Multitask is an approach inductive transfer (using what learned for one assist another problem) and helps improve generalization performance relative each task separately by using domain information contained training signals as bias. Several strategies have been derived within this community under assumption that...

The kernel least-mean-square (KLMS) algorithm is a popular in nonlinear adaptive filtering due to its simplicity and robustness. In filters, the statistics of input linear filter depends on parameters employed. Moreover, practical implementations require finite nonlinearity model order. A Gaussian KLMS has two design parameters, step size bandwidth. Thus, requires analytical models for behavior as function these parameters. This paper studies steady-state transient inputs order model....

Adaptive filtering algorithms operating in reproducing kernel Hilbert spaces have demonstrated superiority over their linear counterpart for nonlinear system identification. Unfortunately, an undesirable characteristic of these methods is that the order filters grows linearly with number input data. This dramatically increases computational burden and memory requirement. A variety strategies based on dictionary learning been proposed to overcome this severe drawback. In literature, there no...

Dynamic system modeling plays a crucial role in the development of techniques for stationary and nonstationary signal processing. Due to inherent physical characteristics systems under investigation, nonnegativity is desired constraint that can usually be imposed on parameters estimate. In this paper, we propose general method identification constraints. We derive so-called nonnegative least-mean-square algorithm (NNLMS) based stochastic gradient descent, analyze its convergence. Experiments...

In this work, we consider multitask learning problems where clusters of nodes are interested in estimating their own parameter vector. Cooperation among is beneficial when the optimal models adjacent have a good number similar entries. We propose fully distributed algorithm for solving problem. The approach relies on minimizing global mean-square error criterion regularized by non-differentiable terms to promote cooperation neighboring clusters. A general diffusion forward-backward splitting...

We consider distributed multitask learning problems over a network of agents where each agent is interested in estimating its own parameter vector, also called task, and the tasks at neighboring are related according to set linear equality constraints. Each possesses convex cost function vector constraints involving vectors agents. propose an adaptive stochastic algorithm based on projection gradient method diffusion strategies order allow optimize individual costs subject all Although...

The multitask diffusion LMS is an efficient strategy to simultaneously infer, in a collaborative manner, multiple parameter vectors. Existing works on problems assume that all agents respond data synchronously. In several applications, may not be able act synchronously because networks can subject sources of uncertainties such as changing topology, random link failures, or turning and off for energy conservation. this paper, we describe model the solution over asynchronous carry out detailed...

Fiber-optic distributed acoustic sensing (DAS) is an emerging technology for vibration measurements with numerous applications in seismic signal analysis, including microseismicity detection, ambient noise tomography, earthquake source characterization, and active seismology. Using laser-pulse techniques, DAS turns (commercial) fiber-optic cables into arrays a spatial sampling density of the order meters time rate up to one thousand Hertz. The versatility enables dense instrumentation...

Hyperspectral image (HSI) super-resolution is commonly used to overcome the hardware limitations of existing hyperspectral imaging systems on spatial resolution. It fuses a low-resolution (LR) HSI and high-resolution (HR) conventional same scene obtain an HR HSI. In this work, we propose method that integrates physical model deep prior information. Specifically, novel, yet effective two-stream fusion network designed serve as regularizer for problem. This problem formulated optimization...

Coupled tensor approximation has recently emerged as a promising approach for the fusion of hyperspectral and multispectral images, reconciling state art performance with strong theoretical guarantees. However, tensor-based approaches previously proposed assume that different observed images are acquired under exactly same conditions. A recent work to accommodate inter-image spectral variability in image problem using matrix factorization-based formulation, but did not account...

Distributed Acoustic Sensing (DAS) is a novel vibration sensing technology that can be employed to detect vehicles and analyse traffic flows using existing telecommunication cables. DAS therefore has great potential in future "smart city" developments, such as real-time incident detection. Though previous studies have considered vehicle detection under relatively light conditions, order for feasible real-world scenarios, algorithms need also perform robustly wide range of conditions. In this...

The prime motivation of our work is to balance the inherent trade-off between resource consumption and accuracy target tracking in wireless sensor networks. Toward this objective, study goes through three phases. First, a cluster-based scheme exploited. At every sampling instant, only one cluster sensors that located proximity activated, whereas other are inactive. To activate most appropriate cluster, we propose nonmyopic rule, which based on not state prediction but also its future...

The tracking of a moving target in wireless sensor network (WSN) requires exact knowledge positions. However, precise information about locations is not always available. Given the observation that series measurements are generated sensors when moves through field, we propose an algorithm exploits these to simultaneously localize detecting and track (SLAT). main difficulties encountered this problem ambiguity locations, unrestricted manner, extremely constrained resources WSNs. Therefore,...

We consider multiagent stochastic optimization problems over reproducing kernel Hilbert spaces. In this setting, a network of interconnected agents aims to learn decision functions, i.e., nonlinear statistical models, that are optimal in terms global convex functional aggregates data across the network, with only access locally and sequentially observed samples. propose solving problem by allowing each agent local regression function while enforcing consensus constraints. use penalized...

Health monitoring is performed on CNES spacecraft using two complementary methods: an automatic Out-Of-Limits (OOL) checking executed a set of critical parameters after each new telemetry reception, and monthly statistical features (daily minimum, mean maximum) another parameters.In this paper we present the limitations system introduce innovative anomaly detection method based machine-learning algorithms, developed during collaborative R&D action between TESA (TElecommunications for Space...

Most works on graph signal processing assume static signals, which is a limitation even in comparison to traditional DSP techniques where signals are modeled as sequences that evolve over time. For broader applicability, it necessary develop able process dynamic or streaming data. Many earlier adaptive networks have addressed problems related this challenge by developing effective strategies particularly well-suited data into graphs. We thus faced with two paradigms: one and sitting the...