Robust and adaptive diffusion-based classification in distributed networks

Abstract

Distributed adaptive signal processing and communication networking are rapidly advancing research areas which enable new and powerful signal processing tasks, e.g., distributed speech enhancement in adverse environments. An emerging new paradigm is that of multiple devices cooperating in multiple tasks (MDMT). This is different from the classical wireless sensor network (WSN) setup, in which multiple devices perform one single joint task. A crucial first step in order to achieve a benefit, e.g., a better node-specific audio signal enhancement, is the common unique labeling of all relevant sources that are observed by the network. This challenging research question can be addressed by designing adaptive data clustering and classification rules based on a set of noisy unlabeled sensor observations. In this paper, two robust and adaptive distributed hybrid classification algorithms are introduced. They consist of a local clustering phase that uses a small part of the data with a subsequent, fully distributed on-line classification phase. The classification is performed by means of distance-based similarity measures. In order to deal with the presence of outliers, the distances are estimated robustly. An extensive simulation-based performance analysis is provided for the proposed algorithms. The distributed hybrid classification approaches are compared to a benchmark algorithm where the error rates are evaluated in dependence of different WSN parameters. Communication cost and computation time are compared for all algorithms under test. Since both proposed approaches use robust estimators, they are, to a certain degree, insensitive to outliers. Furthermore, they are designed in a way that they are applicable to on-line classification problems.

Robust Bayesian cluster enumeration based on the t distribution

“…In real-world applications, the observed data is often subject to heavy tailed noise and outliers [3], [26]- [30] which obscure the true underlying structure of the data. Consequently, cluster enumeration becomes even more challenging when either the data is contaminated by a fraction of outliers or there exist deviations from the distributional assumptions.…”

Gravitational Clustering: A simple, robust and adaptive approach for distributed networks

“…Recently, several distributed algorithms have been proposed, which frame the labelling problem in terms of cluster analysis after extracting source-specific features [6]- [8], [13], [14]. Various methods have been proposed for distributed data clustering, e.g., [14]- [27]. However, a significant drawback of common clustering algorithms is that the number of clusters has to be known a priori.…”

“…The aim of this research is to adaptively estimate the timevarying number of clusters based on a set of streaming-in feature vectors. The proposed method is designed to be 1) adaptive -to a changing number of objects/sources, 2) robust -against outliers in the feature vectors, or in general against unknown non-spherical and possibly heavy tailed distributions of the estimated features, 3) distributed -so as to operate in a decentralized WSN, e.g., based on the diffusion-principle [14], [54], 4) sequential -so that the estimate of the number of clusters is continuously updated for streaming-in data without the need to re-run the entire algorithm, 5) computationally simple -in order to be applicable in a real WSN. Original Contributions: A robust gravitational clustering algorithm is proposed which works for single-node and cooperative in-network clustering.…”

Adaptive diffusion-based track assisted multi-object labeling in distributed camera networks

“…holds the key to the successful cooperation of neighboring nodes in a distributed network. Recently, some generic clustering and classification algorithms have been proposed for distributed sensor networks [4]- [6]. The research question addressed in this paper is to develop a distributed and adaptive multi-object labeling algorithm for a multi-camera network without assuming any form of camera calibration or utilizing a centralized computing unit that fuses all information collected from different cameras.…”