Latent Tree Approximation in Linear Model

TL;DR

This paper introduces a low-complexity method combining EM and Chow-Liu algorithms to learn tree structures from noisy linear model data, suitable for model selection.

Contribution

It proposes a novel approach that integrates EM with Chow-Liu for efficient tree structure learning in linear models with noise.

Findings

Effective in noisy environments

Low computational complexity

Applicable to various linear models

Abstract

We consider the problem of learning underlying tree structure from noisy, mixed data obtained from a linear model. To achieve this, we use the expectation maximization algorithm combined with Chow-Liu minimum spanning tree algorithm. This algorithm is sub-optimal, but has low complexity and is applicable to model selection problems through any linear model.