Encoding large information structures in linear algebra and statistical models

TL;DR

This paper presents methods to encode large information structures in linear algebra and statistical models, enabling faster parameter estimation and dimension reduction for large datasets.

Contribution

It introduces encoding techniques for large data in linear mixed and mixture models to accelerate computation and reduce dimensions.

Findings

Speeded up parameter estimation by a factor based on chosen dimension reduction

Demonstrated encoding reduces sample and feature dimensions effectively

Applicable to linear mixed and mixture models

Abstract

Large information sizes in samples and features can be encoded to speed up the learning of statistical models based on linear algebra and remove unwanted signals. Encoding information can reduce both sample and feature dimension to a smaller representational set. Here two examples are shown on linear mixed models and mixture models speeding up the run time for parameter estimation by a factor defined by the user's choice on dimension reduction (can be linear, quadratic or beyond based on dimension specification).