Search papers, labs, and topics across Lattice.
This paper introduces Streaming Continual Learning (SCL) as a unifying paradigm addressing both rapid adaptation (concept drift) from Streaming Machine Learning (SML) and knowledge retention from Continual Learning (CL). It argues that SCL bridges the CL and SML communities and enables hybrid approaches that combine the strengths of both. Through experiments, the authors demonstrate the limitations of CL and SML in isolation, highlighting the necessity of SCL for scenarios requiring both rapid adaptation and knowledge retention.
Forget juggling CL and SML separately—Streaming Continual Learning (SCL) offers a single framework for rapid adaptation *and* knowledge retention in dynamic environments.
Continual Learning (CL) and Streaming Machine Learning (SML) study the ability of agents to learn from a stream of non-stationary data. Despite sharing some similarities, they address different and complementary challenges. While SML focuses on rapid adaptation after changes (concept drifts), CL aims to retain past knowledge when learning new tasks. After a brief introduction to CL and SML, we discuss Streaming Continual Learning (SCL), an emerging paradigm providing a unifying solution to real-world problems, which may require both SML and CL abilities. We claim that SCL can i) connect the CL and SML communities, motivating their work towards the same goal, and ii) foster the design of hybrid approaches that can quickly adapt to new information (as in SML) without forgetting previous knowledge (as in CL). We conclude the paper with a motivating example and a set of experiments, highlighting the need for SCL by showing how CL and SML alone struggle in achieving rapid adaptation and knowledge retention.
