New metrics for analyzing continual learners
This work addresses the evaluation challenge in continual learning, which is crucial for developing robust AI systems that learn continuously, but it is incremental as it focuses on refining metrics rather than introducing a new learning method.
The paper tackles the problem of evaluating continual learning models by identifying that existing metrics fail to account for increasing task difficulty, which inherently causes performance loss. It proposes new metrics that incorporate task difficulty, demonstrating through experiments on benchmark datasets that these metrics offer new insights into the stability-plasticity trade-off.
Deep neural networks have shown remarkable performance when trained on independent and identically distributed data from a fixed set of classes. However, in real-world scenarios, it can be desirable to train models on a continuous stream of data where multiple classification tasks are presented sequentially. This scenario, known as Continual Learning (CL) poses challenges to standard learning algorithms which struggle to maintain knowledge of old tasks while learning new ones. This stability-plasticity dilemma remains central to CL and multiple metrics have been proposed to adequately measure stability and plasticity separately. However, none considers the increasing difficulty of the classification task, which inherently results in performance loss for any model. In that sense, we analyze some limitations of current metrics and identify the presence of setup-induced forgetting. Therefore, we propose new metrics that account for the task's increasing difficulty. Through experiments on benchmark datasets, we demonstrate that our proposed metrics can provide new insights into the stability-plasticity trade-off achieved by models in the continual learning environment.