Anomaly Detection in Dynamic Networks using Multi-view Time-Series Hypersphere Learning

Teng, Xian; Lin, Yu‐Ru; Wen, Xidao

doi:10.1145/3132847.3132964

Cited by 26 publications

(19 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…25 linear and found that their bilinear network could outperform a CNN and a LSTM, both in terms of computational complexity and precision, in stock price prediction. Furthermore, the concept of bilinear mapping has recently yielded promising results for time series anomaly detection in dynamic maps [55] among others, and we, therefore, evaluate the model in this study. The bilinear map is a function that combines elements of two vector spaces into a third vector space [53].…”

Section: Layermentioning

confidence: 99%

An Experimental Comparison of Anomaly Detection Methods for Collaborative Robot Manipulators

Graabæk¹,

Ancker²,

Christensen³

et al. 2022

Preprint

View full text Add to dashboard Cite

See abstract and datasets [10.5281/zenodo.5849300]<br><br>There exist a large number of methods that can be used for anomaly detection/fault detection in collaborative robots. However, studies on these methods tend to only focus on a single or a couple of such methods, which can make it challenging to gauge their relative merits in specific robot scenarios. In this paper, we conduct a comprehensive comparison of 15 methods for anomaly detection, including methods based on principle component analysis, local outlier factor, and autoencoders. The methods are assessed in a typical pick-and-place application with respect to their capacity to detect a broad range of exogenous anomalies. The results of the study show that several methods perform well, but that their performance profiles differ across the studied anomalies. The results also give an indication of the application characteristics that have the potential to make anomaly detection challenging.

show abstract

Section: Layermentioning

confidence: 99%

An Experimental Comparison of Anomaly Detection Methods for Collaborative Robot Manipulators

Graabæk¹,

Ancker²,

Christensen³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Matrix factorization methods [38,39,41,47] leverage the "lowrank" property of real-world network structures [27] that is mostly represented as overlapping, non-overlapping, or hierarchical community structures [14]. Anomalies break the low-rank property and are thus detectable.…”

Section: Preliminaries and Related Workmentioning

confidence: 99%

F-FADE: Frequency Factorization for Anomaly Detection in Edge Streams

Chang

Sosič

et al. 2020

Preprint

View full text Add to dashboard Cite

Edge streams are commonly used to capture interactions in dynamic networks, such as email, social, or computer networks. The problem of detecting anomalies or rare events in edge streams has a wide range of applications. However, it presents many challenges due to lack of labels, a highly dynamic nature of interactions, and the entanglement of temporal and structural changes in the network. Current methods are limited in their ability to address the above challenges and to efficiently process a large number of interactions.Here, we propose F-FADE, a new approach for detection of anomalies in edge streams, which uses a novel frequency-factorization technique to efficiently model the time-evolving distributions of frequencies of interactions between node-pairs. The anomalies are then determined based on the likelihood of the observed frequency of each incoming interaction. F-FADE is able to handle in an online streaming setting a broad variety of anomalies with temporal and structural changes, while requiring only constant memory. Our experiments on one synthetic and six real-world dynamic networks show that F-FADE achieves state of the art performance and may detect anomalies that previous methods are unable to find.

show abstract

“…In nowadays applications, it is common to observe non-Euclidean data settings, such as timeseries and sequences. A practical solution is to compute a relational representation based on non-Euclidean pairwise dissimilarities between the data points [7]. Consequently, the kernel variants of prototype-based methods are designed by assuming a corresponding implicit mapping to a reproducing kernel Hilbert space (RKHS).…”

Section: Introductionmentioning

confidence: 99%

“…Applying different kernels on the inputs results in a multiple-kernel (MK) representation of the data which might carry non-redundant pieces of information about essential properties of the data [11,7]. Consequently, multiple-kernel learning approaches are designed to find an effective weighted combination of these base kernels that enhances the classification performance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Interpretable Multiple-Kernel Prototype Learning for Discriminative Representation and Feature Selection

Hosseini

Hammer

2019

Proceedings of the 28th ACM International Conference on Information and Knowledge Management

View full text Add to dashboard Cite

Prototype-based methods are of the particular interest for domain specialists and practitioners as they summarize a dataset by a small set of representatives. Therefore, in a classification setting, interpretability of the prototypes is as significant as the prediction accuracy of the algorithm. Nevertheless, the state-of-the-art methods make inefficient trade-offs between these concerns by sacrificing one in favor of the other, especially if the given data has a kernel-based (or multiple-kernel) representation. In this paper, we propose a novel interpretable multiple-kernel prototype learning (IMKPL) to construct highly interpretable prototypes in the feature space, which are also efficient for the discriminative representation of the data. Our method focuses on the local discrimination of the classes in the feature space and shaping the prototypes based on condensed class-homogeneous neighborhoods of data. Besides, IMKPL learns a combined embedding in the feature space in which the above objectives are better fulfilled. When the base kernels coincide with the data dimensions, this embedding results in a discriminative features selection. We evaluate IMKPL on several benchmarks from different domains which demonstrate its superiority to the related state-of-the-art methods regarding both interpretability and discriminative representation. * Preprint of the publication [1], as provided by the authors. The final publication is available at https://dl.acm.org/proceedings.cfm

show abstract

Anomaly Detection in Dynamic Networks using Multi-view Time-Series Hypersphere Learning

Cited by 26 publications

References 29 publications

An Experimental Comparison of Anomaly Detection Methods for Collaborative Robot Manipulators

An Experimental Comparison of Anomaly Detection Methods for Collaborative Robot Manipulators

F-FADE: Frequency Factorization for Anomaly Detection in Edge Streams

Interpretable Multiple-Kernel Prototype Learning for Discriminative Representation and Feature Selection

Contact Info

Product

Resources

About