2020 IEEE International Symposium
on Information Theory
21-26 June 2020 • Los Angeles, California, USA
| Paper ID | L.13.1 | ||
| Paper Title | Active Learning for Classification with Abstention | ||
| Authors | Shubhanshu Shekhar, University of California, San Diego, United States; Mohammad Ghavamzadeh, Facebook AI Research, United States; Tara Javidi, University of California, San Diego, United States | ||
| Session | L.13: Online, Active, and Transfer Learning | ||
| Presentation | Lecture | ||
| Track | Statistics and Learning Theory | ||
| Manuscript | Click here to download the manuscript | ||
| Virtual Presentation | Click here to watch in the Virtual Symposium | ||
| Abstract | We consider the problem of binary classification with the caveat that the learner can abstain from declaring a label incurring a cost $\lambda \in [0,1/2]$ in the process. This is referred to as the problem of binary classification with a fixed-cost of abstention. For this problem, we propose an active learning strategy that constructs a non-uniform partition of the input space and focuses sampling in the regions near the decision boundaries. Our proposed algorithm can work in all the commonly used active learning query models, namely \emph{membership-query}, \emph{pool-based} and \emph{stream-based}. We obtain an upper bound on the excess risk of our proposed algorithm under standard smoothness and margin assumptions and demonstrate its minimax near-optimality by deriving a matching~(modulo poly-logarithmic factors) lower bound. The achieved minimax rates are always faster than the corresponding rates in the passive setting, and furthermore the improvement increases with larger values of the smoothness and margin parameters. | ||
Plan Ahead
2021 IEEE International Symposium on Information Theory
11-16 July 2021 | Melbourne, Victoria, Australia
