윤기중 교수

We performed a follow-up study to address whether high sensitivity C-reactive protein (hs-CRP) levels within the normal range can predict the development of non-alcoholic fatty liver disease (NAFLD) in healthy male subjects. Among15347 male workers between 30 and 59 years old who received annual health check-ups in 2002, a NAFLD-free cohort of 4,138 was followed through December 2009. Alcohol consumption was assessed with a questionnaire. At each visit, abdominal ultrasonography was performed to

In this paper, we revisit the recurrent back-propagation (RBP) algorithm, discuss the conditions under which it applies as well as how to satisfy them in deep neural networks. We show that RBP can be unstable and propose two variants based on conjugate gradient on the normal equations (CG-RBP) and Neumann series (Neumann-RBP). We further investigate the relationship between Neumann-RBP and back propagation through time (BPTT) and its truncated version (TBPTT). Our Neumann-RBP has the same time c

The aim of this study was to evaluate the relationships between job stress and indicators of autonomic nervous system activity in employees of the manufacturing industry. A total of 140 employees from a company that manufactures consumer goods (i.e., diapers and paper towels) were recruited for participation in this study. Job stress was assessed using Karasek's Job Content Questionnaire. Heart rate variability (HRV) was measured using a heart rate monitor, and urinary catecholamines were measur

A fundamental computation for statistical inference and accurate decision-making is to estimate the marginal probabilities or most probable states of task-relevant variables. Probabilistic graphical models can efficiently represent the structure of such complex data, but performing these inferences is generally difficult. Message-passing algorithms, such as belief propagation, are a natural way to disseminate evidence amongst correlated variables while exploiting the graph structure, but these a

In this paper the issue of reasoning about constraints is addressed. A design is derived through the direct manipulation of constraints which narrow down the design space, and through the use of generative mechanisms within the design space. A computer system is described that enables knowledge about spatial constraints to be represented and made operable. The domain under consideration is that of space planning.

Multivariate time series is prevalent in many scientific and industrial domains. Modeling multivariate signals is challenging due to their long-range temporal dependencies and intricate interactions–both direct and indirect. To confront these complexities, we introduce a method of representing multivariate signals as nodes in a graph with edges indicating interdependency between them. Specifically, we leverage graph neural networks (GNN) and attention mechanisms to efficiently learn the underlyi

Graph neural networks (GNNs) have become compelling models designed to perform learning and inference on graph-structured data. However, little work has been done to understand the fundamental limitations of GNNs for scaling to larger graphs and generalizing to out-of-distribution (OOD) inputs. In this paper, we use a random graph generator to systematically investigate how the graph size and structural properties affect the predictive performance of GNNs. We present specific evidence that the a

[This corrects the article DOI: 10.1371/journal.pone.0172666.].