Session 16: High-dimensional machine learning methods

Session title: High-dimensional machine learning methods
Organizer: Annie Qu (UIUC)
Chair: Annie Qu (UIUC)
Time: June 4th, 1:45pm – 3:15pm
Location: VEC 1302/1303

Speech 1: Latent Space Approaches to Community Detection in Dynamic Networks
Speaker: Yuguo Chen  (UIUC)
Abstract: Embedding dyadic data into a latent space has long been a popular approach to modeling networks of all kinds. While clustering has been done using this approach for static networks, we give two methods of community detection within dynamic network data, building upon the distance and projection models previously proposed in the literature. Our proposed approaches capture the time-varying aspect of the data, can model directed or undirected edges, inherently incorporate transitivity and account for each actor’s individual propensity to form edges. We provide Bayesian estimation algorithms, and apply these methods to a ranked dynamic friendship network and world export/import data.

Speech 2: Classification for High-Dimensional Functional Data
Speaker: Taps Maiti (MSU)
Abstract: The functional magnetic resonance imaging (fMRI) records signals coming from different areas in human brains, which show activities and states of brains. This measurements result in high-dimensional time series, and each dimension represents a region in brains. In this work, we propose a classification technique of this kind of high-dimensional time series data while we discuss several competitive approaches. In addition, we present two classification applications to demonstrate the performance of your method. One is an alcoholic condition detection with Electroencephalography (EEG) data collected from electrodes placed on subject’s scalps, and the other is a resting state detection using resting state fMRI data from the OpenfMRI database. In both applications, we compare the performance of our method with some other classification methods.

Speech 3: A unified approach for censored quantile regression
Speaker: Naveen Narisetty (UIUC)
Abstract:  In this talk, I will present a new and unified approach for estimation of quantile regression under censoring of arbitrary types. The proposed method is based on a variation of the data augmentation algorithm and easily adapts to different forms of censoring including doubly censored and interval censored data unlike existing methods which are mostly limited to single censoring. Moreover, the proposed estimators improve on the performance of the best known estimators with singly censored data. Empirical results demonstrating the fine performance of the proposed approach will be presented. This is joint work with Xiaorong Yang and Xuming He.