Continual Learning in Practice
This addresses the need for automated and adaptive ML systems in production, but it is incremental as it focuses on architectural design without new methods.
The paper tackles the problem of building self-maintaining systems for continual learning in evolving data environments, proposing a reference architecture to manage ML models, adapt to shifting distributions, and handle new tasks.
This paper describes a reference architecture for self-maintaining systems that can learn continually, as data arrives. In environments where data evolves, we need architectures that manage Machine Learning (ML) models in production, adapt to shifting data distributions, cope with outliers, retrain when necessary, and adapt to new tasks. This represents continual AutoML or Automatically Adaptive Machine Learning. We describe the challenges and proposes a reference architecture.