Activity Number:
|
242
- Multiple Testing and Feature Selection
|
Type:
|
Contributed
|
Date/Time:
|
Monday, July 30, 2018 : 2:00 PM to 3:50 PM
|
Sponsor:
|
IMS
|
Abstract #328681
|
Presentation
|
Title:
|
Asymptotic Analysis of Large-Scale Multi-Relational Network Through Latent Variable Modeling
|
Author(s):
|
Zhi Wang* and Xueying Tang and Jingchen Liu
|
Companies:
|
Columbia University and Columbia University and Columbia University
|
Keywords:
|
multi-relational network;
latent variable modeling;
KL divergence;
large deviation;
MLE
|
Abstract:
|
We establish consistency results on recovering multi-relational networks driven by a scoring function of low-dimensional latent attributes. We provide a non-asymptotic upper bound for the error probability of estimated distributions in terms of Kullback-Leibler divergence when the latent attributes are estimated by a maximum likelihood estimator or an $L_p$-penalized maximum likelihood estimator. Thanks to the non-asymptotic bound, consistency results hold even when the dimension of the latent space grows slowly with the sample size. Finite sample performance of the estimators is evaluated on simulated and real datasets.
|
Authors who are presenting talks have a * after their name.