Source author record

JungHoon Lee

JungHoon Lee appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Computer Vision Information Theory math.IT cond-mat.mes-hall eess.IV Machine Learning Artificial Intelligence cond-mat.mtrl-sci physics.flu-dyn

Catalog footprint

What is connected

9works

9topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2021arXiv

M2FN: Multi-step Modality Fusion for Advertisement Image Assessment

Assessing advertisements, specifically on the basis of user preferences and ad quality, is crucial to the marketing industry. Although recent studies have attempted to use deep neural networks for this purpose, these studies have not utilized image-related auxiliary attributes, which include embedded text frequently found in ad images. We, therefore, investigated the influence of these attributes on ad image preferences. First, we analyzed large-scale real-world ad log data and, based on our findings, proposed a novel multi-step modality fusion network (M2FN) that determines advertising images likely to appeal to user preferences. Our method utilizes auxiliary attributes through multiple steps in the network, which include conditional batch normalization-based low-level fusion and attention-based high-level fusion. We verified M2FN on the AVA dataset, which is widely used for aesthetic image assessment, and then demonstrated that M2FN can achieve state-of-the-art performance in preference prediction using a real-world ad dataset with rich auxiliary attributes.

preprint2020arXiv

Finding novelty with uncertainty

Medical images are often used to detect and characterize pathology and disease; however, automatically identifying and segmenting pathology in medical images is challenging because the appearance of pathology across diseases varies widely. To address this challenge, we propose a Bayesian deep learning method that learns to translate healthy computed tomography images to magnetic resonance images and simultaneously calculates voxel-wise uncertainty. Since high uncertainty occurs in pathological regions of the image, this uncertainty can be used for unsupervised anomaly segmentation. We show encouraging experimental results on an unsupervised anomaly segmentation task by combining two types of uncertainty into a novel quantity we call scibilic uncertainty.

preprint2020arXiv

Validating uncertainty in medical image translation

Medical images are increasingly used as input to deep neural networks to produce quantitative values that aid researchers and clinicians. However, standard deep neural networks do not provide a reliable measure of uncertainty in those quantitative values. Recent work has shown that using dropout during training and testing can provide estimates of uncertainty. In this work, we investigate using dropout to estimate epistemic and aleatoric uncertainty in a CT-to-MR image translation task. We show that both types of uncertainty are captured, as defined, providing confidence in the output uncertainty estimates.

preprint2014arXiv

Modification of particle-laden near-wall turbulence: effect of Stokes number

Turbulent channel flows laden with particles are investigated using direct numerical simulation with a point-force approximation for small, heavy particles with a diameter smaller than the Kolmogorov length scale of the fluid. The Stokes numbers based on the wall units considered in our study are $St^+=0.5, 5, 35$ and 125. The main purpose of this study is to examine the effect of Stokes number on turbulence modification in a channel. We found that particles with $St^+=0.5$ enhance turbulence by increasing the occurrence of quasistreamwise vortices, while larger-Stokes-number particles attenuate turbulence. When $St^+=0.5$, kinetic energy is transferred from the particles to streamwise fluid velocity fluctuations in the high-speed regions and low-speed streaks, which may increase the instability of the low-speed streaks responsible for the birth of new quasistreamwise vortices. On the other hand, the preferential concentration of larger-Stokes-number particles in low-speed streaks is responsible for turbulence attenuation, and the slow response of the particles to the fluid produces feedback against the fluid velocity associated with quasistreamwise vortices.

preprint2013arXiv

Distributed Detection in Coexisting Large-scale Sensor Networks

This paper considers signal detection in coexisting wireless sensor networks (WSNs). We characterize the aggregate signal and interference from a Poisson random field of nodes and define a binary hypothesis testing problem to detect a signal in the presence of interference. For the testing problem, we introduce the maximum likelihood (ML) detector and simpler alternatives. The proposed mixed-fractional lower order moment (FLOM) detector is computationally simple and close to the ML performance, and robust to estimation errors in system parameters. We also derived asymptotic theoretical performances for the proposed simple detectors. Monte-Carlo simulations are used to supplement our analytical results and compare the performance of the receivers.

preprint2012arXiv

Stochastic Ordering of Interferences in Large-scale Wireless Networks

Stochastic orders are binary relations defined on probability distributions which capture intuitive notions like being larger or being more variable. This paper introduces stochastic ordering of interference distributions in large-scale networks modeled as point process. Interference is the main performance-limiting factor in most wireless networks, thus it is important to understand its statistics. Since closed-form results for the distribution of interference for such networks are only available in limited cases, interference of networks are compared using stochastic orders, even when closed form expressions for interferences are not tractable. We show that the interference from a large-scale network depends on the fading distributions with respect to the stochastic Laplace transform order. The condition for path-loss models is also established to have stochastic ordering between interferences. The stochastic ordering of interferences between different networks are also shown. Monte-Carlo simulations are used to supplement our analytical results.

preprint2011arXiv

Space-Time Coding over Fading Channels with Stable Noise

This paper addresses the performance of space-time coding over fading channels with impulsive noise which is known to accurately capture network interference. We use the symmetric alpha stable noise distribution and adopt two models which assume dependent and independent noise components across receive antennas. We derive pairwise error probability (PEP) of orthogonal space-time block codes (STBC) with a benchmark genie-aided receiver (GAR), or the minimum distance receiver (MDR) which is optimal in the Gaussian case. For general space-time codes we propose a maximum-likelihood (ML) receiver, and its approximation at high signal-to-noise ratio (SNR). The resulting asymptotically optimal receiver (AOR) does not depend on noise parameters and is computationally simple. Monte-Carlo simulations are used to supplement our analytical results and compare the performance of the receivers.

preprint2010arXiv

Dissociation of ssDNA - Single-Walled Carbon Nanotube Hybrids by Watson-Crick Base Pairing

The unwrapping event of ssDNA from the SWNT during the Watson-Crick base paring is investigated through electrical and optical methods, and binding energy calculations. While the ssDNA-metallic SWNT hybrid shows the p-type semiconducting property, the hybridization product recovered metallic properties. The gel electrophoresis directly verifies the result of wrapping and unwrapping events which was also reflected to the Raman shifts. Our molecular dynamics simulations and binding energy calculations provide atomistic description for the pathway to this phenomenon. This nano-physical phenomenon will open up a new approach for nano-bio sensing of specific sequences with the advantages of efficient particle-based recognition, no labeling, and direct electrical detection which can be easily realized into a microfluidic chip format.