Tso-Jung Yen

A majorization–minimization approach to variable selection using spike and slab priors

We develop a method to carry out MAP estimation for a class of Bayesian regression models in which coefficients are assigned with Gaussian-based spike and slab priors. The objective function in the corresponding optimization problem has a Lagrangian form in that regression coefficients are regularized by a mixture of squared

Solving norm constrained portfolio optimization via coordinate-wise descent algorithms

l_2

ClickDiary: Online Tracking of Health Behaviors and Mood

and

Alters as species: Predicting personal network size from contact diaries

l_0

Community Detection in Dynamic Social Networks: A Random Walk Approach

norms. A tight approximation to the

Pareto simplified swarm optimization for grid-computing reliability and service makspan in grid-RMS

l_0

Detecting concurrent mood in daily contact networks: an online participatory cohort study with a diary approach

norm using majorization-minimization techniques is derived, and a coordinate descent algorithm in conjunction with a soft-thresholding scheme is used in searching for the optimizer of the approximate objective. Simulation studies show that the proposed method can lead to more accurate variable selection than other benchmark methods. Theoretical results show that under regular conditions, sign consistency can be established, even when the Irrepresentable Condition is violated. Results on posterior model consistency and estimation consistency, and an extension to parameter estimation in the generalized linear models are provided.

Structured variable selection via prior-induced hierarchical penalty functions

A fast method based on coordinate-wise descent algorithms is developed to solve portfolio optimization problems in which asset weights are constrained by l q norms for 1 ? q ? 2 . The method is first applied to solve a minimum variance portfolio (mvp) optimization problem in which asset weights are constrained by a weighted l 1 norm and a squared l 2 norm. Performances of the weighted norm penalized mvp are examined with two benchmark data sets. When the sample size is not large in comparison with the number of assets, the weighted norm penalized mvp tends to have a lower out-of-sample portfolio variance, lower turnover rate, fewer numbers of active constituents and shortsale positions, but higher Sharpe ratio than the one without such penalty. Several extensions of the proposed method are illustrated; in particular, an efficient algorithm for solving a portfolio optimization problem in which assets are allowed to be chosen grouply is derived.

Learning of hierarchical fuzzy aggregative network using simplified swarm optimization

Background Traditional studies of health behaviors are typically conducted using one-shot, cross-sectional surveys. Thus, participants’ recall bias may undermine the reliability and validity of the data. To capture mood changes and health behaviors in everyday life, we designed an online survey platform, ClickDiary, which helped collect more complete information for comprehensive data analyses. Objective We aim to understand whether daily mood changes are related to one’s personal characteristics, demographic factors, and daily health behaviors. Methods The ClickDiary program uses a Web-based platform to collect data on participants’ health behaviors and their social-contact networks. The name ClickDiary comes from the platform’s interface, which is designed to allow the users to respond to most of the survey questions simply by clicking on the options provided. Participants were recruited from the general population and came from various backgrounds. To keep the participants motivated and interested, the ClickDiary program included a random drawing for rewards. We used descriptive statistics and the multilevel proportional-odds mixed model for our analysis. Results We selected 130 participants who had completed at least 30 days of ClickDiary entries from May 1 to October 31, 2014 as our sample for the study. According to the results of the multilevel proportional-odds mixed model, a person tended to be in a better mood on a given day if he or she ate more fruits and vegetables, took in more sugary drinks, ate more fried foods, showed no cold symptoms, slept better, exercised longer, and traveled farther away from home. In addition, participants were generally in a better mood during the weekend than on weekdays. Conclusions Sleeping well, eating more fruits and vegetables, and exercising longer each day all appear to put one in a better mood. With the online ClickDiary survey, which reduces the recall biases that are common in traditional one-shot surveys, we were able to collect and analyze the daily variations of each subject’s health behaviors and mood status.

Estimating links of a network from time to event data

Abstract Like wildlife species in an ecological system, members within a personal network (or alters) constantly shift and often remain hard to count. Previous studies often estimated the size of such personal networks using information given by a focal person (or ego), who names a list of friends and acquaintances, or someone known or related, that meet certain specified criteria. In a search for alternative methods, we estimate the number of alters using contact diaries that help reveal active and comprehensive interactions, which enable us to predict personal network size from a longitudinal perspective. By exploring contact frequencies between ego and alters, we propose a modeling approach based on species accumulation curves from ecology. Under this approach, the contact frequency between ego and alter often turns out to be a mixture of binomial distributions, and the number of alters with whom ego may make contact in the future is assumed to follow a specified discrete distribution. We estimate the model with the Bayesian nonparametric method, in which the distribution of contact probabilities is assumed to be a mixture of Dirichlet processes. We then demonstrate this approach with a data set containing 48 contact diaries collected over three months and discuss how such an ecological analogy may enrich social network studies.