박태성 교수

We propose spatially-adaptive normalization, a simple but effective layer for synthesizing photorealistic images given an input semantic layout. Previous methods directly feed the semantic layout as input to the network, forcing the network to memorize the information throughout all the layers. Instead, we propose using the input layout for modulating the activations in normalization layers through a spatially-adaptive, learned affine transformation. Experiments on several challenging datasets d

We propose spatially-adaptive normalization, a simple but effective layer for synthesizing photorealistic images given an input semantic layout. Previous methods directly feed the semantic layout as input to the deep network, which is then processed through stacks of convolution, normalization, and nonlinearity layers. We show that this is suboptimal as the normalization layers tend to ``wash away'' semantic information. To address the issue, we propose using the input layout for modulating the

Our results show that intensity-dependent normalization often performs better than global normalization methods, and that linear and nonlinear normalization methods perform similarly. These conclusions are based on analysis of 36 cDNA microarrays of 3,840 genes obtained in an experiment to search for changes in gene expression profiles during neuronal differentiation of cortical stem cells. Simulation studies confirm our findings.

We propose a statistical test procedure based on the ANOVA model to identify genes that have different gene expression profiles among experimental groups in time-course experiments. Especially, we propose a permutation test which does not require the normality assumption. For this test, we use residuals from the ANOVA model only with time-effects. Using this test, we detect genes that have different gene expression profiles among experimental groups. The proposed model is illustrated using cDNA

Deep generative models have become increasingly effective at producing realistic images from randomly sampled seeds, but using such models for controllable manipulation of existing images remains challenging. We propose the Swapping Autoencoder, a deep model designed specifically for image manipulation, rather than random sampling. The key idea is to encode an image with two independent components and enforce that any swapped combination maps to a realistic image. In particular, we encourage the

In image-to-image translation, each patch in the output should reflect the content of the corresponding patch in the input, independent of domain. We propose a straightforward method for doing so -- maximizing mutual information between the two, using a framework based on contrastive learning. The method encourages two elements (corresponding patches) to map to a similar point in a learned feature space, relative to other elements (other patches) in the dataset, referred to as negatives. We expl

We propose GauGAN, a GAN-based image synthesis model that can generate photo-realistic images given an input semantic layout. It is built on spatially-adaptive normalization, a simple but effective normalization layer. Previous methods directly feed the semantic layout as input to the deep network, which is then processed through stacks of convolution, normalization, and non-linearity layers. We show that this is sub-optimal as the normalization layers tend to "wash away" semantic information. T

Unraveling the genetic background of common complex traits is a major goal in modern genetics. In recent years, genome-wide association (GWA) studies have been conducted with large-scale data sets of genetic variants. Most of those studies have relied on single-marker approaches that identify single genetic factors individually and can be limited in considering fully the joint effects of multiple genetic factors on complex traits. Joint identification of multiple genetic factors would be more po

Abstract When categorical outcomes are subject to nonignorable nonresponse, log-linear models may be used to adjust for the nonresponse. The models are fitted to the data in an augmented frequency table in which one index corresponds to whether or not the subject is a respondent. The likelihood function is maximized over pseudo-observed cell frequencies with respect to this log-linear model using an EM algorithm. Each E step of the EM algorithm determines the pseudo-observed cell frequencies, an

The established biomarker for hepatocellular carcinoma (HCC), serum α-fetoprotein (AFP), has suboptimal performance in early disease stages. This study aimed to develop a metabolite panel to differentiate early-stage HCC from cirrhosis. Cross-sectional metabolomic analyses of serum samples were performed for 53 and 47 patients with early HCC and cirrhosis, respectively, and 50 matched healthy controls. Results were validated in 82 and 80 patients with early HCC and cirrhosis, respectively. To re

Liang and Zeger proposed an extension of generalized linear models to the analysis of longitudinal data. Their approach is closely related to quasi-likelihood methods and can handle both normal and non-normal outcome variables such as Poisson or binary outcomes. Their approach, however, has been applied mainly to non-normal outcome variables. This is probably due to the fact that there is a large class of multivariate linear models available for normal outcomes such as growth models and random-e

We aimed to develop a diagnostic model identifying ovarian cancer (OC) from benign ovarian tumors using metagenomic data from serum microbe-derived extracellular vesicles (EVs). We obtained serum samples from 166 patients with pathologically confirmed OC and 76 patients with benign ovarian tumors. For model construction and validation, samples were randomly divided into training and test sets in the ratio 2:1. Isolation of microbial EVs from serum samples of the patients and 16S rDNA amplicon se

Multiple genome-wide and targeted association studies reveal a significant association of variants in the CHRNA5-CHRNA3-CHRNB4 (CHRNA5/A3/B4) gene cluster on chromosome 15 with nicotine dependence. The subjects examined in most of these studies had a European origin. However, considering the distinct linkage disequilibrium patterns in European and other ethnic populations, it would be of tremendous interest to determine whether such associations could be replicated in populations of other ethnic

In this article, we introduce two types of new evaluation measures. First, we develop weighted BA (wBA) that utilizes the quantitative information on the effect size of each multi-locus genotype on a trait. Second, we employ ordinal association measures to assess the performance of MDR classifiers. Simulation studies were conducted to compare the proposed measures with BA, a current measure. Our results showed that the wBA and tau(b) improved the power of MDR in detecting gene-gene interactions.