Abstract: In lifelong learning, tasks (or classes) to be learned arrive sequentially over time in arbitrary order. During training, knowledge from previous tasks can be captured and transferred to subsequent ones to improve sample efficiency. A central idea for lifelong learning is to learn an efficient representation. Then building a classifier on top of it is less expensive than building one from the original input.

In this talk, we propose a lifelong learning algorithm that maintains and refines the internal feature representation. The resulting sample complexity improves significantly on existing bounds. In the setting of linear features, our algorithm is provably efficient and the sample complexity for input dimension $d$, $m$ tasks with $k$ features up to error $\epsilon$ is $\tilde{O}(dk^{1.5}/\epsilon+km/\epsilon)$. We also prove a matching lower bound for any lifelong learning algorithm that uses a single task learner as a black box.

This is based on joint work with Weiyang Liu and Santosh Vempala, appeared in AISTATS 2022.