Active Continual Learning: On Balancing Knowledge Retention and Learnability
This addresses the challenge of efficiently learning from sequential tasks with limited annotations in machine learning, though it is incremental as it builds on existing active and continual learning methods.
The paper tackles the problem of active continual learning (ACL), which combines active learning and continual learning for sequences of tasks with unlabeled data and annotation budgets, revealing a trade-off between knowledge retention and learnability. Experiments show that conditioning active learning on previous annotations improves performance in domain and task-incremental scenarios, but a gap remains in balancing these effects for class-incremental learning.
Acquiring new knowledge without forgetting what has been learned in a sequence of tasks is the central focus of continual learning (CL). While tasks arrive sequentially, the training data are often prepared and annotated independently, leading to the CL of incoming supervised learning tasks. This paper considers the under-explored problem of active continual learning (ACL) for a sequence of active learning (AL) tasks, where each incoming task includes a pool of unlabelled data and an annotation budget. We investigate the effectiveness and interplay between several AL and CL algorithms in the domain, class and task-incremental scenarios. Our experiments reveal the trade-off between two contrasting goals of not forgetting the old knowledge and the ability to quickly learn new knowledge in CL and AL, respectively. While conditioning the AL query strategy on the annotations collected for the previous tasks leads to improved task performance on the domain and task incremental learning, our proposed forgetting-learning profile suggests a gap in balancing the effect of AL and CL for the class-incremental scenario.