Real-world data streams continuously evolve, requiring Machine Learning models to adapt to changing distributions while preserving previously acquired knowledge. Some streaming scenarios, such as sensor data streams, also exhibit temporal dependencies. This setting is defined as Evolving Streaming Time Series (ESTS).

Streaming Machine Learning (SML) and Continual Learning (CL) address complementary aspects of these challenges, but neither fully captures their interplay. Streaming Continual Learning (SCL) is an emerging research paradigm gaining recognition and bringing together the SML and CL communities. It bridges this gap by focusing on the current concept while building a cumulative knowledge repository, from which relevant knowledge can be selected, reused, and modulated.

This talk explores the motivations behind SCL, focusing on its temporal extension, which integrates methodologies from Time Series Analysis to address ESTS. The vision is illustrated through two neural architectures, namely Dynamic Continuous Progressive Neural Networks (DYNcPNN) and MAGIC Net, exploring strategies for knowledge preservation, reuse, and adaptive architectural expansion.

Finally, the talk discusses open challenges and future research directions, highlighting opportunities and calls to action for the research community.

Bio

Federico Giannini is a Postdoctoral Researcher at Politecnico di Milano, where he earned his PhD in Computer Science and Engineering. His research focuses on Streaming Continual Learning (SCL), an emerging paradigm bridging Streaming Machine Learning and Continual Learning, with particular emphasis on developing novel architectures embodying the SCL vision. He has actively contributed to the definition and development of this paradigm and is strongly committed to building a scientific community around it, co-organizing international events at ESANN, AAAI, and ECML PKDD. Before pursuing his academic career, he worked as a Data Scientist at a major insurance company, gaining practical experience in real-world Machine Learning applications. He is also actively involved in teaching, delivering lectures in undergraduate, graduate, and postgraduate programs, as well as corporate training.

Updated: