arXiv subjects

Kwangsik Nho

Publications and source records attributed to Kwangsik Nho.

11 recordsLinked to original sources

arXiv2021

Alzheimer's Disease Diagnosis via Deep Factorization Machine Models

The current state-of-the-art deep neural networks (DNNs) for Alzheimer's Disease diagnosis use different biomarker combinations to classify patients, but do not allow extracting knowledge about the interactions of biomarkers. However, to improve our understanding of the disease, it is paramount to extract such knowledge from the learned model. In this paper, we propose a Deep Factorization Machine model that combines the ability of DNNs to learn complex relationships and the ease of interpretability of a linear model. The proposed model has three parts: (i) an embedding layer to deal with sparse categorical data, (ii) a Factorization Machine to efficiently learn pairwise interactions, and (iii) a DNN to implicitly model higher order interactions. In our experiments on data from the Alzheimer's Disease Neuroimaging Initiative, we demonstrate that our proposed model classifies cognitive normal, mild cognitive impaired, and demented patients more accurately than competing models. In addition, we show that valuable knowledge about the interactions among biomarkers can be obtained.

cs.LG↗

arXiv2019

Deep Learning in Alzheimer's disease: Diagnostic Classification and Prognostic Prediction using Neuroimaging Data

Deep learning has shown outstanding performance in identifying intricate structures in complex high-dimensional data, especially in the domain of computer vision. The application of deep learning to early detection and automated classification of Alzheimer's disease (AD) has recently gained considerable attention, as rapid progress in neuroimaging techniques has generated large-scale multimodal neuroimaging data. A systematic review of publications using deep learning approaches and neuroimaging data for diagnostic classification of AD was performed. A PubMed and Google Scholar search was used to identify deep learning papers on AD published between January 2013 and July 2018. These papers were reviewed, evaluated, and classified by algorithm and neuroimaging type, and the findings were summarized. Of 16 studies meeting full inclusion criteria, 4 used a combination of deep learning and traditional machine learning approaches, and 12 used only deep learning approaches. The combination of traditional machine learning for classification and stacked auto-encoder (SAE) for feature selection produced accuracies of up to 98.8% for AD classification and 83.7% for prediction of conversion from mild cognitive impairment (MCI), a prodromal stage of AD, to AD. Deep learning approaches, such as convolutional neural network (CNN) or recurrent neural network (RNN), that use neuroimaging data without preprocessing for feature selection have yielded accuracies of up to 96.0% for AD classification and 84.2% for MCI conversion prediction. The best classification performance was obtained when multimodal neuroimaging and fluid biomarkers were combined. AD research that uses deep learning is still evolving, improving performance by incorporating additional hybrid data types, increasing transparency with explainable approaches that add knowledge of specific disease-related features and mechanisms.

eess.IV↗

arXiv2006

Finite-temperature properties of quasi-2D Bose-Einstein condensates

Using the finite-temperature path integral Monte Carlo method, we investigate dilute, trapped Bose gases in a quasi-two dimensional geometry. The quantum particles have short-range, s-wave interactions described by a hard-sphere potential whose core radius equals its corresponding scattering length. The effect of both the temperature and the interparticle interaction on the equilibrium properties such as the total energy, the density profile, and the superfluid fraction is discussed. We compare our accurate results with both the semi-classical approximation and the exact results of an ideal Bose gas. Our results show that for repulsive interactions, (i) the minimum value of the aspect ratio, where the system starts to behave quasi-two dimensionally, increases as the two-body interaction strength increases, (ii) the superfluid fraction for a quasi-2D Bose gas is distinctly different from that for both a quasi-1D Bose gas and a true 3D system, i.e., the superfluid fraction for a quasi-2D Bose gas decreases faster than that for a quasi-1D system and a true 3D system with increasing temperature, and shows a stronger dependence on the interaction strength, (iii) the superfluid fraction for a quasi-2D Bose gas lies well below the values calculated from the semi-classical approximation, and (iv) the Kosterlitz-Thouless transition temperature decreases as the strength of the interaction increases.

cond-mat.stat-mech↗

arXiv2005

Bose-Einstein condensation of trapped atoms with dipole interactions

The path integral Monte Carlo method is used to simulate dilute trapped Bose gases and to investigate the equilibrium properties at finite temperatures. The quantum particles have a long-range dipole-dipole interaction and a short-range s-wave interaction. Using an anisotropic pseudopotential for the long-range dipolar interaction and a hard-sphere potential for the short-range s-wave interaction, we calculate the energetics and structural properties as a function of temperature and the number of particles. Also, in order to determine the effects of dipole-dipole forces and the influence of the trapping field on the dipolar condensate, we use two cylindrically symmetric harmonic confinements (a cigar-shaped trap and a disk-shaped trap). We find that the net effect of dipole-dipole interactions is governed by the trapping geometry. For a cigar-shaped trap, the net contribution of dipolar interactions is attractive and the shrinking of the density profiles is observed. For a disk-shaped trap, the net effect of long-range dipolar forces is repulsive and the density profiles expand.

cond-mat.stat-mech↗

arXiv2005

Scaling of the thermal resistivity of $^4$He in restricted geometries

The thermal resistivity and its scaling function in quasi-2D $^4$He systems are studied by Monte Carlo and spin-dynamics simulation. We use the classical 3D XY model on $L\times L\times H$ lattices with $L\gg H$, applying open boundary condition along the $H$ direction and periodic boundary conditions along the $L$ directions. A hybrid Monte Carlo algorithm is adopted to efficiently deal with the critical slowing down and to produce initial states for time integration. Fourth-order Suzuki-Trotter decomposition method of exponential operators is used to solve numerically the coupled equations of motion for each spin. The thermal conductivity is calculated by a dynamic current-current correlation function. Our results show the validity of the finite-size scaling theory, and the calculated scaling function agrees well with the available experimental results for slabs using the temperature scale and the thermal resistivity scale as free fitting parameters.

cond-mat.other↗

arXiv2004

Bose-Einstein Condensation Temperature of a Homogeneous Weakly Interacting Bose Gas : PIMC study

Using a finite-temperature Path Integral Monte Carlo simulation (PIMC) method and finite-size scaling, we have investigated the interaction-induced shift of the phase transition temperature for Bose-Einstein condensation of homogeneous weakly interacting Bose gases in three dimensions, which is given by a proposed analytical expression $T_{c} = T_{c}^{0}\{1 + c_{1}an^{1/3}+[c'_{2}\ln(an^{1/3})+c''_{2}]a^{2}n^{2/3} +O(a^{3}n)\}$, where $T_{c}^{0}$ is the critical temperature for an ideal gas, $a$ is the s-wave scattering length, and $n$ is the number density. We have used smaller number densities and more time slices than in the previous PIMC simulations [Gruter {\it et al.}, Phys. Rev. Lett. {\bf 79}, 3549 (1997)] in order to understand the difference in the value of the coefficient $c_{1}$ between their results and the (apparently) other reliable results in the literature. Our results show that $\{(T_{c}-T_{c}^{0})/T_{c}^{0}\}/(an^{1/3})$ depends strongly on the interaction strength $an^{1/3}$ while the previous PIMC results are considerably flatter and smaller than our results. We obtain $c_{1}$ = 1.32 $\pm$ 0.14, in agreement with results from recent Monte Carlo methods of three-dimensional O(2) scalar $ϕ^{4}$ field theory and variational perturbation theory.

cond-mat.stat-mech↗

arXiv2003

Heat capacity scaling function for confined superfluids

We study the specific heat scaling function of superfluids confined in cubic geometry and in parallel-plate (film) geometry with open boundary conditions (BC) along the finite dimensions using Monte Carlo simulation. For the case of cubic geometry for the superfluid order parameter we apply open BC in all three directions. We also calculated the specific heat scaling function for the parallel-plate confinement using open BC along the finite dimension and periodic BC along the other two dimensions and we find it to be very close to the earlier calculated using Dirichlet instead of open BC. We find that the specific heat scaling function is significantly different for the two different geometries studied. In addition, we generally find that the scaling function for a fixed given geometry when calculated with open BC is quite close to that calculated with Dirichlet BC, while when calculated with periodic BC is quite different. Our results for both the scaling functions obtained for the parallel-plate geometry and for cubic geometry with open BC along the finite dimensions are in very good agreement with the recent very high quality experimental measurements with no free parameters.

cond-mat.str-el↗

arXiv2002

Commensurate-incommensurate transitions in quantum films: Submonolayer molecular hydrogen on graphite

We have used the Path Integral Monte Carlo method to simulate a monolayer of molecular hydrogen on graphite above 1/3 submonolayer coverage. We find that at low temperature and as the coverage increases the system undergoes a series of transformations starting from the $\sqrt{3} \times \sqrt{3}$ commensurate solid near 1/3 coverage. First, a phase is formed which is characterized by a uniaxially compressed incommensurate solid with additional mass density modulations along the same direction which can be viewed as an ordered domain-wall solid with a characteristic domain-wall distance which depends on the surface coverage. At low temperature and higher coverage there is a transition to an incommensurate solid which is rotated relative to the substrate commensurate lattice. As a function of temperature the domain-wall ordering first melts into a fluid of domain-walls and at higher temperature the solid melts into a uniform fluid. Regardless of the large amplitude of quantum fluctuations, these phase transitions are analogous to those in classical monolayer films. Our calculated values for the surface coverage and temperature where these transitions occur, the calculated structure factors and specific heat are in general agreement with the available experimental results with no adjustable parameteters.

cond-mat.str-el↗

arXiv2002

Spin-dynamics simulations of the triangular antiferromagnetic XY model

Using Monte Carlo and spin-dynamics methods, we have investigated the dynamic behavior of the classical, antiferromagnetic XY model on a triangular lattice with linear sizes $L \leq 300$. The temporal evolutions of spin configurations were obtained by solving numerically the coupled equations of motion for each spin using fourth-order Suzuki-Trotter decompositions of exponential operators. From space- and time-displaced spin-spin correlation functions and their space-time Fourier transforms we obtained the dynamic structure factor $S({\bf q},w)$ for momentum ${\bf q}$ and frequency $ω$. Below $T_{KT}$(Kosterlitz-Thouless transition), both the in-plane ($S^{xx}$) and the out-of-plane ($S^{zz}$) components of $S({\bf q},ω)$ exhibit very strong and sharp spin-wave peaks. Well above $T_{KT}$, $S^{xx}$ and $S^{zz}$ apparently display a central peak, and spin-wave signatures are still seen in $S^{zz}$. In addition, we also observed an almost dispersionless domain-wall peak at high $ω$ below $T_{c}$(Ising transition), where long-range order appears in the staggered chirality. Above $T_{c}$, the domain-wall peak disappears for all $q$. The lineshape of these peaks is captured reasonably well by a Lorentzian form. Using a dynamic finite-size scaling theory, we determined the dynamic critical exponent $z$ = 1.002(3). We found that our results demonstrate the consistency of the dynamic finite-size scaling theory for the characteristic frequeny $ω_{m}$ and the dynamic structure factor $S({\bf q},ω)$ itself.

cond-mat.stat-mech↗

arXiv2001

Scaling of thermal conductivity of helium confined in pores

We have studied the thermal conductivity of confined superfluids on a bar-like geometry. We use the planar magnet lattice model on a lattice $H\times H\times L$ with $L \gg H$. We have applied open boundary conditions on the bar sides (the confined directions of length $H$) and periodic along the long direction. We have adopted a hybrid Monte Carlo algorithm to efficiently deal with the critical slowing down and in order to solve the dynamical equations of motion we use a discretization technique which introduces errors only $O((δt)^6)$ in the time step $δt$. Our results demonstrate the validity of scaling using known values of the critical exponents and we obtained the scaling function of the thermal resistivity. We find that our results for the thermal resistivity scaling function are in very good agreement with the available experimental results for pores using the temp

cond-mat.stat-mech↗

arXiv1999

Nematic phase of the two-dimensional electron gas in a magnetic field

The two dimensional electron gas (2DEG) in moderate magnetic fields in ultra-clean AlAs-GaAs heterojunctions exhibits transport anomalies suggestive of a compressible, anisotropic metallic state. Using scaling arguments and Monte Carlo simulations, we develop an order parameter theory of an electron nematic phase. The observed temperature dependence of the resistivity anisotropy behaves like the orientational order parameter if the transition to the nematic state occurs at a finite temperature, $T_c \sim 65 mK$, and is slightly rounded by a small background microscopic anisotropy. We propose a light scattering experiment to measure the critical susceptibility.

cond-mat.mes-hall↗