Talk by Dr. Patrick Schäfer:
"Unleashing the Potential of Unsupervised Time Series Analytics: Recent Advances and Breakthroughs"
31 March 2023 - 10:30
.Dr. Patrick Schäfer, Postdoctoral Researcher and Lecturer of Computer Science at the Humboldt University of Berlin
His main research interests include scalable TS analytics, including supervised tasks, such as classification, and unsupervised tasks, such as motif discovery and segmentation. He is a member of the editorial board of the DAMI journal, and a core developer of sktime.
In recent years, the usage of low-cost, high-resolution sensors has witnessed a surge, with applications in various domains, such as mobile devices, manufacturing monitoring, environmental and medical surveillance, and more. These sensors produce vast amounts of unlabeled, real-valued sequences, also referred to as time series (TS). Despite the abundance of data, the research has primarily focused on supervised techniques for analyzing TS, with a focus on classification and deep learning methods.
However, the field of unsupervised time series analytics (UTSA) offers a broad spectrum of tools to extract valuable information from time series data. This includes primitives like Chains, Discords, Motifs, or Change Points. These enable us to gain insights into the inherent statistical properties and temporal patterns of the underlying processes and enhance our understanding of the generated data. In this talk, I will present our recent advances in UTSA, specifically the detection of approximately repeated frequent patterns (Motiflets) and segmentation (ClaSP, ClaSS).
Bldg. 50.34, room 301