ICIEA 2026 Workshop

Workshop 3: From Traditional Deep Learning to Foundation Models for Time Series Data Analytics

Speakers:

Speaker 1: Zhenghua Chen

Title: Senior Lecturer (Associate Professor)

Affiliation: University of Glasgow

Summary of workshop:

Time series data play a central role in modern industrial systems, energy systems, and predictive health management. However, traditional deep learning methods for time series analysis often suffer from limited labeled data, strong domain shifts across operating conditions, and poor generalization in real-world deployments. This workshop provides an overview of recent advances in time series data analytics, starting from task-specific deep learning models and moving towards self-supervised learning, domain adaptation, and efficient model deployment. It highlights how robust representations can be learned from largely unlabeled data and transferred across different machines, environments, and tasks. The workshop further explores emerging directions toward time series foundation models, including the integration of large language models (LLMs) and knowledge-enhanced graph learning for PHM applications. Recent research results demonstrate how foundation-model-style pretraining can improve generalization, data efficiency, and adaptability in complex industrial scenarios. This workshop aims to offer a clear and accessible perspective on the evolution from traditional deep learning to foundation models for time series data analytics.