Skip to yearly menu bar Skip to main content


Poster

On Complexity of Teaching a Family of Linear Behavior Cloning Learners

Shubham Bharti · Stephen Wright · Adish Singla · Jerry Zhu

West Ballroom A-D #6603
[ ]
Wed 11 Dec 11 a.m. PST — 2 p.m. PST

Abstract: We study optimal teaching for a family of Behavior Cloning learners that learn using a linear hypothesis class. In this setup, a knowledgeable teacher can demonstrate a dataset of state and action tuples, and is required to teach an optimal policy to an entire family of BC learners using smallest possible dataset. We analyse the linear family and design a novel teaching algorithm called `TIE' that achieves the instance optimal teaching complexity for the entire family. However, we show that this problem is NP-hard for action spaces with $|\mathcal{A}| > 2$ and provide an approximation algorithm with a $\log(|\mathcal{A}| - 1)$ guarantee on the optimal teaching size. We present empirical results to demonstrate the effectiveness of our TIE algorithm compared to various baselines in different real-world teaching environments.

Live content is unavailable. Log in and register to view live content