Exploiting a Zoo of Checkpoints for Unseen Tasks

Jiaji Huang, Qiang Qiu, Kenneth Church

Keywords: task space, generalization, Gaussian process, meta learning

TL;DR: This paper characterizes the relationship among tasks, and then further proposes a greedy method to select checkpoints that can generalize well to new tasks.

Abstract: There are so many models in the literature that it is difficult for practitioners to decide which combinations are likely to be effective for a new task. This paper attempts to address this question by capturing relationships among checkpoints published on the web. We model the space of tasks as a Gaussian process. The covariance can be estimated from checkpoints and unlabeled probing data. With the Gaussian process, we can identify representative checkpoints by a maximum mutual information criterion. This objective is submodular. A greedy method identifies representatives that are likely to "cover'' the task space. These representatives generalize to new tasks with superior performance. Empirical evidence is provided for applications from both computational linguistics as well as computer vision.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

Code: https://github.com/baidu-research/task_space