Variational Approximation for Efficient Estimation of Latent Variable Models

• In Event: Dimensionality

Abstract

This study introduces variational approximation as an efficient alternative to Hamiltonian Monte Carlo (HMC) for latent variable model (e.g., CFA and IRT models) parameter estimation. Through a simulation study, we demonstrate that Variational Approximation offers precise results in high-dimensional, small-sample settings while significantly reducing computational time.

Authors

• Jihang Chen, BOSTON COLLEGE

• Zhushan Mandy Li, BOSTON COLLEGE