Title: Regularized Babington-Smith ranking model
Authors: Jong-june Jeon - University of Seoul (Korea, South) [presenting]
Abstract: The aim is to present the regularization method for the Babington-Smith model that derives academically important ranking models. By regularizing parameters of the model, we construct a continuum class of ranking models that bridges Babington-Smith model and Bradley-Terry-Mallows model. Through the regularization we can account for an unusual characteristic such as intransitivity of preferences in the model. We also propose the computational algorithm based on constrastive divergence to estimate the parameters in our model and prove its convergence property of the algorithm. We investigate the model selection problem in our model and provide the results of numerical simulations.