Jing Zhou

Renmin University of China | RUC · School of Statistics

In this work, we propose a progressive principal component analysis (PPCA) method for compressing deep convolutional neural networks. The proposed method starts with a prespecified layer and progressively moves on to the final output layer. For each target layer, PPCA conducts kernel principal component analysis for the estimated kernel weights. Th...

We study herein an autoregressive model with spatially correlated error terms and missing data. A logistic regression model with completely observed covariates is used to model the missingness mechanism. An autoregressive model is used to accommodate time series dependence, and a spatial error model is used to capture spatial dependence. To estimat...

The multivariate GARCH (MGARCH) model is popular for analyzing financial time series data. However, statistical inferences for MGARCH models are quite challenging, owing to the high dimension issue. To overcome this difficulty , we propose a network GARCH model that uses information derived from an appropriately defined network structure. This decr...

A novel function of live streaming is that viewers can send paid gifts to broadcasters. In addition, viewers can engage with broadcasters by sending danmaku, a type of comment scrolled across the screen in real time. This paper investigates the role of viewers’ social interaction in paid gifting on live streaming platforms. We argue that viewer-vie...

Spatial autocorrelation is a parameter of importance for network data analysis. To estimate spatial autocorrelation, maximum likelihood has been popularly used. However, its rigorous implementation requires the whole network to be observed. This is practically infeasible if network size is huge (e.g., Facebook, Twitter, Weibo, WeChat, etc.). In tha...

Deep neural network (DNN) models often involve high-dimensional features. In most cases, these high-dimensional features can be decomposed into two parts: a low-dimensional factor and residual features with much-reduced variability and inter-feature correlation. This decomposition has several interesting theoretical implications for DNN training. B...

Background: The aim of the study was to explore the outcomes of wedge resection on patients with early-stage lung adenocarcinoma (LUAD) and further identify potential prognostic factors for these patients. Methods: A retrospective cohort of 190 patients (99 solitary LUAD and 91 multifocal LUAD) undergone wedge resection from October 2014 to Sept...

Discrete choice model is probably one of the most popularly used statistical methods in practice. The common feature of this model is that it considers the behavioral factors of a person and the assumption of independent individuals. However, this widely accepted assumption seems problematic because human beings do not live in isolation. They inter...

Naive Bayes (NB) is one of the most popular classification methods. It is particularly useful when the dimension of the predictor is high and data are generated independently. In the meanwhile, social network data are becoming increasingly accessible, due to the fast development of various social network services and websites. By contrast, data gen...

In social network analysis, link prediction is a problem of fundamental importance. How to conduct a comprehensive and principled link prediction, by taking various network structure information into consideration, is of great interest. To this end, we propose here a dynamic logistic regression method. Specifically, we assume that one has observed...