Statistical estimation requires unbounded memory
Computation
2011-10-28 v2
Abstract
We investigate the existence of bounded-memory consistent estimators of various statistical functionals. This question is resolved in the negative in a rather strong sense. We propose various bounded-memory approximations, using techniques from automata theory and stochastic processes. Some questions of potential interest are raised for future work.
Keywords
Cite
@article{arxiv.0907.1997,
title = {Statistical estimation requires unbounded memory},
author = {Leonid and Kontorovich},
journal= {arXiv preprint arXiv:0907.1997},
year = {2011}
}
Comments
this is an old version, with a mistake in the proof of Thm. 6.1. Please see my homepage for an updated version