LOCATION: UPV/EHU Donosti, Faculty of Computer Science, room 3.1 and Online

Link to the session here

Abstract
Time Series anomaly detection has emerged as a prominent and active research area in recent times. Traditionally, machine learning methods have approached this task from an unsupervised standpoint. These models are trained to learn the normal behavior of data and identify anomalies by quantifying their level of abnormality in the inference phase. However, these approaches often struggle with generalization, as they tend to overfit to the normal data patterns observed during training. Consequently, they fail to effectively detect anomalies in new samples that exhibit slight variations in properties and patterns. To address this challenge, several novel contributions have been made, leveraging self-supervised learning techniques. Self-supervised learning is an unsupervised methodology that seeks to capture the underlying structure of data by predicting what is already known about it. This presentation offers a concise introduction to the application of self-supervised learning-based approaches aimed at enhancing the performance of anomaly detection frameworks for time series data.