Title: Improving generalization in data-driven models with task-specific knowledge

Abstract:
With the rise of the over-parameterized deep learning models and massive datasets, many have started advocating towards minimizing the amount of prior knowledge added to a learning model. Ironically, the traditional machine learning community advocated for exactly the opposite. Whereas the latter assumes knowledge of all the rules/priors needed to do well at a task, the former assumes access to perfect data, that is, a large amount of clean, i.i.d sampled data. In real-world scenarios, neither of these assumptions is entirely true. We consider four arbitrary domains as examples of some of these scenarios, namely, point cloud completion (with distribution shift), visual dialog (dataset size/bias issues), meta-rl for control (noisy, high variance and sparse training signal) and poaching prediction task (unstructured dataset with skew, noise and distribution shift). Using these datasets, we show that data and priors are meant to complement each other in machine learning models and it’s important to think of them jointly on a task to task basis.

Committee:
Katia Sycara (advisor)
David Held
Wenhao Luo
Aditya Murli