English

How many faces can be recognized? Performance extrapolation for multi-class classification

Machine Learning 2016-06-17 v1 Computer Vision and Pattern Recognition Information Theory Machine Learning math.IT

Abstract

The difficulty of multi-class classification generally increases with the number of classes. Using data from a subset of the classes, can we predict how well a classifier will scale with an increased number of classes? Under the assumption that the classes are sampled exchangeably, and under the assumption that the classifier is generative (e.g. QDA or Naive Bayes), we show that the expected accuracy when the classifier is trained on kk classes is the k1k-1st moment of a \emph{conditional accuracy distribution}, which can be estimated from data. This provides the theoretical foundation for performance extrapolation based on pseudolikelihood, unbiased estimation, and high-dimensional asymptotics. We investigate the robustness of our methods to non-generative classifiers in simulations and one optical character recognition example.

Keywords

Cite

@article{arxiv.1606.05228,
  title  = {How many faces can be recognized? Performance extrapolation for multi-class classification},
  author = {Charles Y. Zheng and Rakesh Achanta and Yuval Benjamini},
  journal= {arXiv preprint arXiv:1606.05228},
  year   = {2016}
}

Comments

Submitted to NIPS 2016

R2 v1 2026-06-22T14:27:08.216Z