Transformers for Time Series - Is the New State of the Art Approaching?

Explore the emerging potential of transformers in time series analysis through this insightful conference talk. Embark on a comprehensive tour of well-known transformer architectures, examining their applications and challenges in the time series domain. Discover why adapting NLP-focused architectures to time series data presents unique difficulties. Gain valuable insights into significant open-source implementations and research papers, including Informer and Spacetimeformer. Delve into a practical implementation of latency prediction in a Kubernetes cluster, comparing it with current state-of-the-art methods. Analyze the pros and cons of these approaches and understand the current landscape of transformer applications in time series analysis. While avoiding deep technical details, this presentation provides a broad overview of the field's progress and future potential.

Transformes for Time Series: Is the New State of the Art (SOA) Approaching? - Ezequiel Lanza, Intel