IRT and FA are Neural Networks: a Friendly Introduction

• In Event: Network Models

Abstract

We describe how to configure a Neural Network to enable it to estimate the parameters of either Item Response Theory or Factor Analysis models. For this purpose, we utilize the architecture of Variational Autoencoders, which is based on Variational Inference. Real data examples are used for benchmarking.

Authors

• Denis Federiakin, Johannes Gutenberg University Mainz

• Lidia Dobria, WILBUR WRIGHT COLLEGE

• Kevin Shenavai, Johannes Gutenberg University Mainz

• Susanne Schmidt

• Olga Zlatkin-Troitschanskaia, JOHANNES GUTENBERG-UNIVERSITY