More Supervision, Less Computation: Statistical-Computational Tradeoffs in Weakly Supervised Learning
Abstract
We consider the weakly supervised binary classification problem where the labels are randomly flipped with probability . Although there exist numerous algorithms for this problem, it remains theoretically unexplored how the statistical accuracies and computational efficiency of these algorithms depend on the degree of supervision, which is quantified by . In this paper, we characterize the effect of by establishing the information-theoretic and computational boundaries, namely, the minimax-optimal statistical accuracy that can be achieved by all algorithms, and polynomial-time algorithms under an oracle computational model. For small , our result shows a gap between these two boundaries, which represents the computational price of achieving the information-theoretic boundary due to the lack of supervision. Interestingly, we also show that this gap narrows as increases. In other words, having more supervision, i.e., more correct labels, not only improves the optimal statistical accuracy as expected, but also enhances the computational efficiency for achieving such accuracy.
Cite
@article{arxiv.1907.06257,
title = {More Supervision, Less Computation: Statistical-Computational Tradeoffs in Weakly Supervised Learning},
author = {Xinyang Yi and Zhaoran Wang and Zhuoran Yang and Constantine Caramanis and Han Liu},
journal= {arXiv preprint arXiv:1907.06257},
year = {2019}
}
Comments
This work has been published in NeurIPS 2016. The first three authors contribute equally