An On-Line Agglomerative Clustering Method for Nonstationary Data

Abstract

An on-line agglomerative clustering algorithm for non-stationary data is described. Three issues are addressed. The rst regards the temporal aspects of the data. The clustering of stationary data by the proposed algorithm is comparable to the other popular algorithms tested (batch and on-line). The second issue addressed is the number of clusters required to represent the data. The algorithm provides an e cient framework to determine the natural number of clusters given the scale of the problem. Finally, the proposed algorithm implicitly minimizes the local distortion { a measure which t a k es into account clusters with relatively small mass.In contrast, most existing on-line clustering methods assume stationarity o f t h e d a t a . W h e n used to cluster non-stationary data these methods fail to generate a good representation. Moreover, most current algorithms are computationally intensive when determining the correct number of clusters. These algorithms tend to neglect clusters of small mass due to their minimization of the global distortion (Energy).

Unsupervised online activity discovery using temporal behaviour assumption

“…Thus, as a basis for our method we used an online agglomerative clustering method introduced in [13]. The algorithm is fairly simple: (i) for each new point it moves the closest centroid towards it; (ii) it merges the two closest centroids; (iii) the new point becomes also a centroid.…”

“…Our method addresses these limitations and builds upon [13] and [14] by incorporating the temporal assumption of the human behaviour, i.e., samples closer in time are likely to form a cluster that represents an activity. The proposed method can handle large number of activities (17 and 33 in our experiments).…”

“…It is based on the online clustering of Guedalia et al [13] and Zhang et al [14], and improves upon by: i) exploiting the human behaviour characteristics and discovering clusters that are continuous in time and last for a predefined minimum duration; ii) including a mechanism to regulate sensitivity to brief outliers; iii) discovering multiple clusters overlapping in time which allows to better model activities comprising deviations from nominal behaviour; iv) being unbounded (open-ended) and not requiring specification of the number of clusters to discover; v) being online and memory and time efficient by not keeping the data samples in memory and having an execution time which does not grow with the number of activities discovered.…”

A Multi-Modal Panoramic Attentional Model for Robots and Applications

“…Each message is associated with an angle corresponding to the azimuthal direction of the origin of the sound from the environment. In order to identify the number of speakers, the messages are clustered in real-time using the azimuthal angle as the clustering attribute, each cluster corresponding to a speaker and the center of the cluster corresponding to the azimuthal directional location of the speaker (Guedalia et al, 1998). During interaction, it is quite natural for the speakers to move about in the environment and this includes entirely disappearing from the zone of interaction.…”

Rapid Online Learning of Objects in a Biologically Motivated Recognition Architecture

“…the growing neural gas [3]. Another related work not dealing with incremental learning, but with clustering of nonstationary or changing datasets was proposed by [5].…”