On the Kolmogorov Complexity of Binary Classifiers

01/28/2022

∙

by   Samuel Epstein, et al.

∙

0

∙

share

We provide tight upper and lower bounds on the expected minimum Kolmogorov complexity of binary classifiers that are consistent with labeled samples. The expected size is not more than complexity of the target concept plus the conditional entropy of the labels given the sample.