arXiv subjects

Jihoon Kim

Publications and source records attributed to Jihoon Kim.

34 records · Page 2Linked to original sources

arXiv2023

Performance Comparison of Design Optimization and Deep Learning-based Inverse Design

Surrogate model-based optimization has been increasingly used in the field of engineering design. It involves creating a surrogate model with objective functions or constraints based on the data obtained from simulations or real-world experiments, and then finding the optimal solution from the model using numerical optimization methods. Recent advancements in deep learning-based inverse design methods have made it possible to generate real-time optimal solutions for engineering design problems, eliminating the requirement for iterative optimization processes. Nevertheless, no comprehensive study has yet closely examined the specific advantages and disadvantages of this novel approach compared to the traditional design optimization method. The objective of this paper is to compare the performance of traditional design optimization methods with deep learning-based inverse design methods by employing benchmark problems across various scenarios. Based on the findings of this study, we provide guidelines that can be taken into account for the future utilization of deep learning-based inverse design. It is anticipated that these guidelines will enhance the practical applicability of this approach to real engineering design problems.

math.OC↗

arXiv2023

Superconductor Pb$_{10-x}$Cu$_x$(PO$_4$)$_6$O showing levitation at room temperature and atmospheric pressure and mechanism

A material called LK-99, a modified-lead apatite crystal structure with the composition Pb$_{10-x}$Cu$_x$(PO$_4$)$_6$O ($0.9<x<1.1$), has been synthesized using the solid-state method. The material exhibits the Ohmic metal characteristic of Pb(6s1) above its superconducting critical temperature, $T_c$, and the levitation phenomenon as Meissner effect of a superconductor at room temperature and atmospheric pressure below $T_c$. A LK-99 sample shows $T_c$ above 126.85$^\circ$C (400 K). We analyze that the possibility of room-temperature superconductivity in this material is attributed to two factors: the first being the volume contraction resulting from an insulator-metal transition achieved by substituting Pb with Cu, and the second being on-site repulsive Coulomb interaction enhanced by the structural deformation in the one-dimensional(D) chain (Cu$^{2+}$-O$_{1/2}$-Cu$^{2+}$ along the c-axis) structure owing to superconducting condensation at $T_c$. The mechanism of the room-temperature $T_c$ is discussed by 1-D BR-BCS theory.

cond-mat.supr-con↗

arXiv2023

FLAME: Free-form Language-based Motion Synthesis & Editing

Text-based motion generation models are drawing a surge of interest for their potential for automating the motion-making process in the game, animation, or robot industries. In this paper, we propose a diffusion-based motion synthesis and editing model named FLAME. Inspired by the recent successes in diffusion models, we integrate diffusion-based generative models into the motion domain. FLAME can generate high-fidelity motions well aligned with the given text. Also, it can edit the parts of the motion, both frame-wise and joint-wise, without any fine-tuning. FLAME involves a new transformer-based architecture we devise to better handle motion data, which is found to be crucial to manage variable-length motions and well attend to free-form text. In experiments, we show that FLAME achieves state-of-the-art generation performances on three text-motion datasets: HumanML3D, BABEL, and KIT. We also demonstrate that editing capability of FLAME can be extended to other tasks such as motion prediction or motion in-betweening, which have been previously covered by dedicated models.

cs.CV↗

arXiv2022

Simulation-guided Beam Search for Neural Combinatorial Optimization

Neural approaches for combinatorial optimization (CO) equip a learning mechanism to discover powerful heuristics for solving complex real-world problems. While neural approaches capable of high-quality solutions in a single shot are emerging, state-of-the-art approaches are often unable to take full advantage of the solving time available to them. In contrast, hand-crafted heuristics perform highly effective search well and exploit the computation time given to them, but contain heuristics that are difficult to adapt to a dataset being solved. With the goal of providing a powerful search procedure to neural CO approaches, we propose simulation-guided beam search (SGBS), which examines candidate solutions within a fixed-width tree search that both a neural net-learned policy and a simulation (rollout) identify as promising. We further hybridize SGBS with efficient active search (EAS), where SGBS enhances the quality of solutions backpropagated in EAS, and EAS improves the quality of the policy used in SGBS. We evaluate our methods on well-known CO benchmarks and show that SGBS significantly improves the quality of the solutions found under reasonable runtime assumptions.

cs.LG↗

arXiv2022

Learning Joint Representation of Human Motion and Language

In this work, we present MoLang (a Motion-Language connecting model) for learning joint representation of human motion and language, leveraging both unpaired and paired datasets of motion and language modalities. To this end, we propose a motion-language model with contrastive learning, empowering our model to learn better generalizable representations of the human motion domain. Empirical results show that our model learns strong representations of human motion data through navigating language modality. Our proposed method is able to perform both action recognition and motion retrieval tasks with a single model where it outperforms state-of-the-art approaches on a number of action recognition benchmarks.

cs.CV↗

arXiv2022

Polarization control of electron injection and acceleration in the plasma by a self steepening laser pulse

We describe an interplay between two injection mechanism of background electrons into an evolving plasma bubble behind an intense laser pulse: one due to the overall bubble expansion, and another due to its periodic undulation. The two mechanisms occur simultaneously when an intense laser pulse propagating inside a plasma forms a shock like steepened front. Periodic undulations of the plasma bubble along the laser propagation path can either inhibit or conspire with electron injection due to bubble expansion. We show that Carrier Envelope Phase (CEP) controlled plasma bubble undulation induced by the self steepening laser pulse produces a unique electron injector: Expanding Phase controlled Undulating Bubble (EPUB). The longitudinal structure of the electron bunch injected by the EPUB can be controlled by laser polarization and power, resulting in high-charge (multiple nano Coulombs) high-current (tens of kilo Amperes) electron beams with ultra short (femtosecond scale) temporal structure. Generation of high energy betatron radiation with polarization and CEP controlled energy spectrum and angular distribution is analyzed as a promising application of EPUB produced beams.

physics.acc-ph↗

arXiv2022

Conditional Motion In-betweening

Motion in-betweening (MIB) is a process of generating intermediate skeletal movement between the given start and target poses while preserving the naturalness of the motion, such as periodic footstep motion while walking. Although state-of-the-art MIB methods are capable of producing plausible motions given sparse key-poses, they often lack the controllability to generate motions satisfying the semantic contexts required in practical applications. We focus on the method that can handle pose or semantic conditioned MIB tasks using a unified model. We also present a motion augmentation method to improve the quality of pose-conditioned motion generation via defining a distribution over smooth trajectories. Our proposed method outperforms the existing state-of-the-art MIB method in pose prediction errors while providing additional controllability.

cs.CV↗

arXiv2022

WAND-PIC: an accelerated three-dimensional quasi-static particle-in-cell code

We introduce a quasi-static particle-in-cell (PIC) code -- WAND-PIC -- which does not suffer from some of the common limitations of many quasi-static PICs, such as the need for a predictor-corrector method in solving electromagnetic fields. We derive the field equations under quasi-static (QS) approximation and find the explicit form of the "time" derivative of the transverse plasma current. After that, equations for the magnetic fields can be solved exactly without using the predictor-corrector method. Algorithm design and code structure are thus greatly simplified. With the help of explicit quasi-static equations and our adaptive step size, plasma bubbles driven by the large beam charges can be simulated efficiently without suffering from the numerical instabilities associated with the predictor-corrector method. In addition, WAND-PIC is able to simulate the sophisticated interactions between high-frequency laser fields and beam particles through the method of sub-cycling. Comparisons between the WAND-PIC and a first-principle full PIC code (VLPL) are presented. WAND-PIC is open-source, fully three-dimensional, and parallelized with the in-house multigrid solver. Scalability, time complexity, and parallelization efficiency up to thousands of cores are also discussed in this work.

physics.plasm-ph↗

arXiv2021

Sub-femtosecond wakefield injector and accelerator based on an undulating plasma bubble controlled by a laser phase

We demonstrate that a long-propagating plasma bubble executing undulatory motion can be produced in the wake of two co-propagating laser pulses: a near-single-cycle injector and a multi-cycle driver. When the undulation amplitude exceeds the analytically derived threshold, highly-localized injections of plasma electrons into the bubble are followed by their long-distance acceleration. While the locations of the injection regions are controlled by the carrier-envelope phase (CEP) of the injector pulse, the mono-energetic spectrum of the accelerated sub-femtosecond high-charge electron bunches is shown to be nearly CEP-independent

physics.plasm-ph↗

arXiv2020

VaB-AL: Incorporating Class Imbalance and Difficulty with Variational Bayes for Active Learning

Active Learning for discriminative models has largely been studied with the focus on individual samples, with less emphasis on how classes are distributed or which classes are hard to deal with. In this work, we show that this is harmful. We propose a method based on the Bayes' rule, that can naturally incorporate class imbalance into the Active Learning framework. We derive that three terms should be considered together when estimating the probability of a classifier making a mistake for a given sample; i) probability of mislabelling a class, ii) likelihood of the data given a predicted class, and iii) the prior probability on the abundance of a predicted class. Implementing these terms requires a generative model and an intractable likelihood estimation. Therefore, we train a Variational Auto Encoder (VAE) for this purpose. To further tie the VAE with the classifier and facilitate VAE training, we use the classifiers' deep feature representations as input to the VAE. By considering all three probabilities, among them especially the data imbalance, we can substantially improve the potential of existing methods under limited data budget. We show that our method can be applied to classification tasks on multiple different datasets -- including one that is a real-world dataset with heavy data imbalance -- significantly outperforming the state of the art.

cs.LG↗

arXiv2019

Summary Level Training of Sentence Rewriting for Abstractive Summarization

As an attempt to combine extractive and abstractive summarization, Sentence Rewriting models adopt the strategy of extracting salient sentences from a document first and then paraphrasing the selected ones to generate a summary. However, the existing models in this framework mostly rely on sentence-level rewards or suboptimal labels, causing a mismatch between a training objective and evaluation metric. In this paper, we present a novel training signal that directly maximizes summary-level ROUGE scores through reinforcement learning. In addition, we incorporate BERT into our model, making good use of its ability on natural language understanding. In extensive experiments, we show that a combination of our proposed model and training procedure obtains new state-of-the-art performance on both CNN/Daily Mail and New York Times datasets. We also demonstrate that it generalizes better on DUC-2002 test set.

cs.CL↗

arXiv2018

Constrained energy minimization based upscaling for coupled flow and mechanics

In this paper, our aim is to present (1) an embedded fracture model (EFM) for coupled flow and mechanics problem based on the dual continuum approach on the fine grid and (2) an upscaled model for the resulting fine grid equations. The mathematical model is described by the coupled system of equation for displacement, fracture and matrix pressures. For a fine grid approximation, we use the finite volume method for flow problem and finite element method for mechanics. Due to the complexity of fractures, solutions have a variety of scales, and fine grid approximation results in a large discrete system. Our second focus in on constructing the upscaled coarse grid poroelasticity model for fractured media. Our upscaled approach is based on the nonlocal multicontinuum (NLMC) upscaling for coupled flow and mechanics problem, which involves computations of local basis functions via an energy minimization principle. This concept allows a systematic upscaling for processes in the fractured porous media, and provides an effective coarse scale model whose degrees of freedoms have physical meaning. We obtain a fast and accurate solver for the poroelasticity problem on a coarse grid and, at the same time, derive a novel upscaled model. We present numerical results for the two dimensional model problem.

math.NA↗

arXiv2018

The Impact of Automatic Pre-annotation in Clinical Note Data Element Extraction - the CLEAN Tool

Objective. Annotation is expensive but essential for clinical note review and clinical natural language processing (cNLP). However, the extent to which computer-generated pre-annotation is beneficial to human annotation is still an open question. Our study introduces CLEAN (CLinical note rEview and ANnotation), a pre-annotation-based cNLP annotation system to improve clinical note annotation of data elements, and comprehensively compares CLEAN with the widely-used annotation system Brat Rapid Annotation Tool (BRAT). Materials and Methods. CLEAN includes an ensemble pipeline (CLEAN-EP) with a newly developed annotation tool (CLEAN-AT). A domain expert and a novice user/annotator participated in a comparative usability test by tagging 87 data elements related to Congestive Heart Failure (CHF) and Kawasaki Disease (KD) cohorts in 84 public notes. Results. CLEAN achieved higher note-level F1-score (0.896) over BRAT (0.820), with significant difference in correctness (P-value < 0.001), and the mostly related factor being system/software (P-value < 0.001). No significant difference (P-value 0.188) in annotation time was observed between CLEAN (7.262 minutes/note) and BRAT (8.286 minutes/note). The difference was mostly associated with note length (P-value < 0.001) and system/software (P-value 0.013). The expert reported CLEAN to be useful/satisfactory, while the novice reported slight improvements. Discussion. CLEAN improves the correctness of annotation and increases usefulness/satisfaction with the same level of efficiency. Limitations include untested impact of pre-annotation correctness rate, small sample size, small user size, and restrictedly validated gold standard. Conclusion. CLEAN with pre-annotation can be beneficial for an expert to deal with complex annotation tasks involving numerous and diverse target data elements.

cs.CL↗

arXiv2017

Evolutionary models of red supergiants: Evidence for a metallicity-dependent mixing length and implications for Type IIP supernova progenitors

Recent studies on the temperatures of red supergiants (RSGs) in the local universe provide us with an excellent observational constraint on RSG models. We calibrate the mixing length parameter by comparing model predictions with the empirical RSG temperatures in Small and Large Magellanic Clouds, Milky Way, and M31, which are inferred from the TiO band and the spectral energy distribution (SED). Although our RSG models are computed with the MESA code, our result may be applied to other stellar evolution codes, including the BEC and TWIN codes. We find evidence that the mixing length increases with increasing metallicity for both cases where the TiO and SED temperatures of RSGs are used for the calibration. Together with the recent finding of a similar correlation in low-mass red giants by Tayar et al, this implies that the metallicity dependence of the mixing length is a universal feature in post-main sequence stars of both low and high masses. Our result implies that typical Type IIP supernova (SN IIP) progenitors with initial masses of $\sim 10 - 16~M_\odot$ have a radius range of $400 R_\odot \lesssim R \lesssim 800 R_\odot$ regardless of metallicity. As an auxiliary result of this study, we find that the hydrogen-rich envelope mass of SN IIP progenitors for a given initial mass is predicted to be largely independent of metallicity if the Ledoux criterion with slow semiconvection is adopted, while the Schwarzschild models predict systematically more massive hydrogen-rich envelopes for lower metallicity.

astro-ph.SR↗

arXiv2013

Systematic Determination of Absolute Absorption Cross-section of Individual Carbon Nanotubes

Determination of optical absorption cross-section is always among the central importance of understanding a material. However its realization on individual nanostructures, such as carbon nanotubes, is experimentally challenging due to the small extinction signal using conventional transmission measurements. Here we develop a technique based on polarization manipulation to enhance the sensitivity of single-nanotube absorption spectroscopy by two-orders of magnitude. We systematically determine absorption cross-section over broad spectral range at single-tube level for more than 50 chirality-defined single-walled nanotubes. Our data reveals chirality-dependent one-dimensional photo-physics through the behaviours of exciton oscillator strength and lifetime. We also establish an empirical formula to predict absorption spectrum of any nanotube, which provides the foundation to determine quantum efficiencies in important photoluminescence and photovoltaic processes.

cond-mat.mes-hall↗

arXiv2002

Thermochemistry of MgB2 Thin Film Synthesis

We have investigated the thermodynamic and kinetic barriers involved in the synthesis of MgB2 films. This work refines our initial conjectures predicting optimal MgB2 thin film growth conditions as a consequence of the unusually large kinetic barrier to MgB2 decomposition. The small Mg sticking coefficient at temperatures greater than 300C prevents high temperature synthesis with traditional vacuum growth methods. However, as a result of the large kinetic barrier to MgB2 decomposition, in-situ thermal processing can be used to enhance the crystallinity and the superconductivity of MgB2 films. We used these methods to produce MgB2 thin films with relatively high transition temperatures (37K) by pulsed laser deposition (PLD).

cond-mat.supr-con↗