On Estimating Non-uniform Density Distributions using N Nearest Neighbors

TL;DR

This paper introduces a method for estimating non-uniform density distributions using N nearest neighbors, normalizing hypersphere volumes and employing Legendre polynomials to improve accuracy in sparse data scenarios.

Contribution

It proposes a novel normalization technique and polynomial modeling approach to enhance density estimation from N nearest neighbors in arbitrary dimensions.

Findings

Effective bias-variance tradeoff achieved

Normalization improves density estimation accuracy

Method applicable to high-dimensional, sparse data

Abstract

We consider density estimators based on the nearest neighbors method applied to discrete point distibutions in spaces of arbitrary dimensionality. If the density is constant, the volume of a hypersphere centered at a random location is proportional to the expected number of points falling within the hypersphere radius. The distance to the $N$-th nearest neighbor alone is then a sufficient statistic for the density. In the non-uniform case the proportionality is distorted. We model this distortion by normalizing hypersphere volumes to the largest one and expressing the resulting distribution in terms of the Legendre polynomials. Using Monte Carlo simulations we show that this approach can be used to effectively address the tradeoff between smoothing bias and estimator variance for sparsely sampled distributions.