BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260415T215125EDT-3047uUi1Zm@132.216.98.100
DTSTAMP:20260416T015125Z
DESCRIPTION:Abstract\n\nTime-Series Anomaly Detection (TSAD)\, the task of 
 identifying patterns that deviate from expected behavior\, is critical in 
 domains such as e-commerce\, cybersecurity\, predictive maintenance\, and 
 healthcare. Despite substantial progress\, TSAD remains challenging due to
  the complexity of time-series signals\, the diversity of anomaly types\, 
 and the scarcity of high-quality labeled data. This thesis addresses these
  challenges through three complementary contributions.\n\nFirst\, the fiel
 d lacks a systematic understanding of how emerging techniques such as grap
 h modeling and self-supervised learning (SSL) can be leveraged for anomaly
  detection. Existing surveys often overlook the unique challenges of TSAD\
 , leaving researchers without a roadmap to guide future work. To address t
 his gap\, we present the first comprehensive surveys on Graph-based TSAD (
 G-TSAD)\, a novel perspective on modeling time-series data using graph str
 uctures for the task of TSAD\, and on Self-Supervised Learning for Anomaly
  Detection (SSL-AD)\, which demonstrates how proxy tasks can assist TSAD i
 n obtaining robust representations from unlabeled data. These surveys high
 light methodological advances\, practical limitations\, and provide an out
 look on promising future directions for TSAD.\n\nSecond\, while graph-base
 d approaches have recently been introduced to capture spatial relationship
 s across sensors in multivariate sensory systems\, they often overlook fin
 e-grained local structures\, such as sub-graphs\, that can be critical for
  detecting anomalies. To address this\, we propose EEG-CGS\, a novel contr
 astive and generative SSL framework for anomaly detection in complex senso
 ry systems. EEG-CGS incorporates local structural patterns into graph repr
 esentations while requiring no anomaly labels during training. This design
  improves robustness in multivariate TSAD and demonstrates strong performa
 nce in detecting anomalous sensors and regions.\n\nFinally\, a key challen
 ge in unsupervised TSAD lies in the assumption that training data are pure
 ly normal\, which is rarely valid in practice due to distribution shifts o
 r labeling errors. Such contamination causes unsupervised methods to overf
 it and misclassify anomalies encountered during training. To address this\
 , we introduce TSAD-C\, a novel framework that incorporates graph represen
 tations and diffusions models\, to capture both long-term temporal and spa
 tial dependencies in time series\, while explicitly handling contamination
 . Furthermore\, unlike existing TSAD approaches benchmarked on small\, cur
 ated datasets with simplistic anomalies\, this thesis advances TSAD toward
 s frameworks that generalize to complex\, real-world scenarios and detect 
 richer anomaly types\, from local signal deviations to sensor- and region-
 level failures\, with direct applications in clinical and industrial domai
 ns.\n
DTSTART:20260220T180000Z
DTEND:20260220T200000Z
LOCATION:Room 603\, McConnell Engineering Building\, CA\, QC\, Montreal\, H
 3A 0E9\, 3480 rue University
SUMMARY:PhD defence of Thi Kieu Khanh Ho – Time-Series Anomaly Detection wi
 th Graphs
URL:https://www.mcgill.ca/ece/channels/event/phd-defence-thi-kieu-khanh-ho-
 time-series-anomaly-detection-graphs-371278
END:VEVENT
END:VCALENDAR