arXiv subjects

Chanwook Park

Publications and source records attributed to Chanwook Park.

18 recordsLinked to original sources

arXiv2026

DiffATS: Diffusion in Aligned Tensor Space

Direct diffusion modeling of high-resolution spatiotemporal fields is computationally challenging. Parameter-efficient primitives address this by representing high-dimensional data with a compact set of parameters. In this paper, we construct data-dependent tensor primitives without pretrained compression autoencoders. Our construction starts from Tucker decomposition, which captures low-rank multilinear structure through a core tensor and mode-wise factors. However, Tucker factors are non-unique: the same tensor can be represented by different rotated factors, which complicates generative modeling. We address this issue with orthogonal Procrustes (OP) alignment. Specifically, we select medoid anchor matrices from the data and align the factor matrices to resolve the gauge ambiguity. This yields matrix Grassmannian primitives and tensor Grassmannian primitives that are compact, data-adaptive, and directly decodable by explicit multilinear reconstruction. Theoretically, we prove that the proposed primitive maps are homeomorphisms between low-rank tensors and their corresponding primitive spaces, certifying that the representations are non-degenerate and topologically faithful. Building on these primitives, we propose *Diffusion in Aligned Tensor Space* (DiffATS), a generative framework that trains diffusion models directly on aligned tensor primitives. Across images, videos, and PDE solutions, DiffATS achieves strong unconditional and conditional generation performance while compressing original data by $3.9\times$ to $210\times$, without relying on any pretrained deep compression autoencoders.

cs.LG↗

arXiv2026

Bayesian Interpolating Neural Network (B-INN): a scalable and reliable Bayesian model for large-scale physical systems

Neural networks and machine learning models for uncertainty quantification suffer from limited scalability and poor reliability compared to their deterministic counterparts. In industry-scale active learning settings, where generating a single high-fidelity simulation may require days or weeks of computation and produce data volumes on the order of gigabytes, they quickly become impractical. This paper proposes a scalable and reliable Bayesian surrogate model, termed the Bayesian Interpolating Neural Network (B-INN). The B-INN combines high-order interpolation theory with tensor decomposition and alternating direction algorithm to enable effective dimensionality reduction without compromising predictive accuracy. We theoretically show that the function space of a B-INN is a subset of that of Gaussian processes, while its Bayesian inference exhibits linear complexity, $\mathcal{O}(N)$, with respect to the number of training samples. Numerical experiments demonstrate that B-INNs can be from 20 times to 10,000 times faster with a robust uncertainty estimation compared to Bayesian neural networks and Gaussian processes. These capabilities make B-INN a practical foundation for uncertainty-driven active learning in large-scale industrial simulations, where computational efficiency and robust uncertainty calibration are paramount.

math.NA↗

arXiv2025

Large language model-empowered next-generation computer-aided engineering

Software development has entered a new era where large language models (LLMs) now serve as general-purpose reasoning engines, enabling natural language interaction and transformative applications across diverse domains. This paradigm is now extending into computer-aided engineering (CAE). Recent applications of LLMs in CAE have successfully automated routine tasks, including CAD model generation and FEM simulations. Nevertheless, these contributions, which primarily serve to reduce manual labor, are often insufficient for addressing the significant computational challenges posed by large-scale, high-dimensional systems. To this aim, we first introduce the concept of LLM-empowered CAE agent, where LLMs act as autonomous collaborators that plan, execute, and adapt CAE workflows. Then, we propose an LLM-empowered CAE agent for data-free model order reduction (MOR), a powerful yet underused approach for ultra-fast large-scale parametric analysis due to the intrusive nature and labor-intensive redevelopment of solvers. LLMs can alleviate this barrier by automating derivations, code restructuring, and implementation, making intrusive MOR both practical and broadly accessible. To demonstrate feasibility, we present an LLM-empowered CAE agent for solving ultra-large-scale space-parameter-time (S-P-T) physical problems using Tensor-decomposition-based A Priori Surrogates (TAPS). Our results show that natural language prompts describing parametric partial differential equations (PDEs) can be translated into efficient solver implementations, substantially reducing human effort while producing high-fidelity reduced-order models. Moreover, LLMs can synthesize novel MOR solvers for unseen cases such as nonlinear and high-dimensional parametric problems based on their internal knowledge base. This highlights the potential of LLMs to establish the foundation for next-generation CAE systems.

cs.CE↗

arXiv2025

A Convolutional Hierarchical Deep-learning Neural Network (C-HiDeNN) Framework for Non-linear Finite Element Analysis

We present a framework for the Convolutional Hierarchical Deep-learning Neural Network (C-HiDeNN) tailored for nonlinear finite element analysis. Building upon the structured foundation of HiDeNN, C-HiDeNN introduces a convolution operator to enhance numerical approximation. A distinctive feature of C-HiDeNN is its higher-order accurate approximation achieved through an expanded set of parameters, such as the polynomial order 'p,' dilation parameter 'a,' patch size 's,' and nodal position 'X'. These parameters function as the functional equivalents of weights and biases within each C-HiDeNN patch. In addition, C-HiDeNN can be selectively applied to regions requiring high resolution to adaptively improve local prediction accuracy. To demonstrate the effectiveness of this framework, we provide numerical examples in the context of nonlinear finite element analysis. The results show that our approach achieves significantly higher accuracy than conventional Finite Element Method (FEM) while substantially reducing computational costs.

math.NA↗

arXiv2025

Explainable Hierarchical Deep Learning Neural Networks (Ex-HiDeNN)

Data-driven science and computation have advanced immensely to construct complex functional relationships using trainable parameters. However, efficiently discovering interpretable and accurate closed-form expressions from complex dataset remains a challenge. The article presents a novel approach called Explainable Hierarchical Deep Learning Neural Networks or Ex-HiDeNN that uses an accurate, frugal, fast, separable, and scalable neural architecture with symbolic regression to discover closed-form expressions from limited observation. The article presents the two-step Ex-HiDeNN algorithm with a separability checker embedded in it. The accuracy and efficiency of Ex-HiDeNN are tested on several benchmark problems, including discerning a dynamical system from data, and the outcomes are reported. Ex-HiDeNN generally shows outstanding approximation capability in these benchmarks, producing orders of magnitude smaller errors compared to reference data and traditional symbolic regression. Later, Ex-HiDeNN is applied to three engineering applications: a) discovering a closed-form fatigue equation, b) identification of hardness from micro-indentation test data, and c) discovering the expression for the yield surface with data. In every case, Ex-HiDeNN outperformed the reference methods used in the literature. The proposed method is built upon the foundation and published works of the authors on Hierarchical Deep Learning Neural Network (HiDeNN) and Convolutional HiDeNN. The article also provides a clear idea about the current limitations and future extensions of Ex-HiDeNN.

cs.LG↗

arXiv2025

INN-FF: A Scalable and Efficient Machine Learning Potential for Molecular Dynamics

The ability to accurately model interatomic interactions in large-scale systems is fundamental to understanding a wide range of physical and chemical phenomena, from drug-protein binding to the behavior of next-generation materials. While machine learning interatomic potentials (MLIPs) have made it possible to achieve ab initio-level accuracy at significantly reduced computational cost, they still require very large training datasets and incur substantial training time and expense. In this work, we propose the Interpolating Neural Network Force Field (INN-FF), a novel framework that merges interpolation theory and tensor decomposition with neural network architectures to efficiently construct molecular dynamics potentials from limited quantum mechanical data. Interpolating Neural Networks (INNs) achieve comparable or better accuracy than traditional multilayer perceptrons (MLPs) while requiring orders of magnitude fewer trainable parameters. On benchmark datasets such as liquid water and rMD17, INN-FF not only matches but often surpasses state-of-the-art accuracy by an order of magnitude, while achieving significantly lower error when trained on smaller datasets. These results suggest that INN-FF offers a promising path toward building efficient and scalable machine-learned force fields.

cond-mat.mtrl-sci↗

arXiv2025

Tensor-decomposition-based A Priori Surrogate (TAPS) modeling for ultra large-scale simulations

A data-free, predictive scientific AI model, Tensor-decomposition-based A Priori Surrogate (TAPS), is proposed for tackling ultra large-scale engineering simulations with significant speedup, memory savings, and storage gain. TAPS can effectively obtain surrogate models for high-dimensional parametric problems with equivalent zetta-scale ($10^{21}$) degrees of freedom (DoFs). TAPS achieves this by directly obtaining reduced-order models through solving governing equations with multiple independent variables such as spatial coordinates, parameters, and time. The paper first introduces an AI-enhanced finite element-type interpolation function called convolution hierarchical deep-learning neural network (C-HiDeNN) with tensor decomposition (TD). Subsequently, the generalized space-parameter-time Galerkin weak form and the corresponding matrix form are derived. Through the choice of TAPS hyperparameters, an arbitrary convergence rate can be achieved. To show the capabilities of this framework, TAPS is then used to simulate a large-scale additive manufacturing process as an example and achieves around 1,370x speedup, 14.8x memory savings, and 955x storage gain compared to the finite difference method with $3.46$ billion spatial degrees of freedom (DoFs). As a result, the TAPS framework opens a new avenue for many challenging ultra large-scale engineering problems, such as additive manufacturing and integrated circuit design, among others.

cs.CE↗

arXiv2025

Interpolating Neural Network-Tensor Decomposition (INN-TD): a scalable and interpretable approach for large-scale physics-based problems

Deep learning has been extensively employed as a powerful function approximator for modeling physics-based problems described by partial differential equations (PDEs). Despite their popularity, standard deep learning models often demand prohibitively large computational resources and yield limited accuracy when scaling to large-scale, high-dimensional physical problems. Their black-box nature further hinders the application in industrial problems where interpretability and high precision are critical. To overcome these challenges, this paper introduces Interpolating Neural Network-Tensor Decomposition (INN-TD), a scalable and interpretable framework that has the merits of both machine learning and finite element methods for modeling large-scale physical systems. By integrating locally supported interpolation functions from finite element into the network architecture, INN-TD achieves a sparse learning structure with enhanced accuracy, faster training/solving speed, and reduced memory footprint. This makes it particularly effective for tackling large-scale high-dimensional parametric PDEs in training, solving, and inverse optimization tasks in physical problems where high precision is required.

cs.CE↗

arXiv2024

IANUS: Integrated Accelerator based on NPU-PIM Unified Memory System

Accelerating end-to-end inference of transformer-based large language models (LLMs) is a critical component of AI services in datacenters. However, diverse compute characteristics of end-to-end LLM inference present challenges as previously proposed accelerators only address certain operations or stages (e.g., self-attention, generation stage, etc.). To address the unique challenges of accelerating end-to-end inference, we propose IANUS -- Integrated Accelerator based on NPU-PIM Unified Memory System. IANUS is a domain-specific system architecture that combines a Neural Processing Unit (NPU) with a Processing-in-Memory (PIM) to leverage both the NPU's high computation throughput and the PIM's high effective memory bandwidth. In particular, IANUS employs a unified main memory system where the PIM memory is used both for PIM operations and for NPU's main memory. The unified main memory system ensures that memory capacity is efficiently utilized and the movement of shared data between NPU and PIM is minimized. However, it introduces new challenges since normal memory accesses and PIM computations cannot be performed simultaneously. Thus, we propose novel PIM Access Scheduling that manages normal memory accesses and PIM computations through workload mapping and scheduling across the PIM and the NPU. Our detailed simulation evaluations show that IANUS improves the performance of GPT-2 by 6.2$\times$ and 3.2$\times$, on average, compared to the NVIDIA A100 GPU and the state-of-the-art accelerator. As a proof-of-concept, we develop a prototype of IANUS with a commercial PIM, NPU, and an FPGA-based PIM controller to demonstrate the feasibility of IANUS.

cs.AR↗

arXiv2024

Convolutional Hierarchical Deep Learning Neural Networks-Tensor Decomposition (C-HiDeNN-TD): a scalable surrogate modeling approach for large-scale physical systems

A common trend in simulation-driven engineering applications is the ever-increasing size and complexity of the problem, where classical numerical methods typically suffer from significant computational time and huge memory cost. Methods based on artificial intelligence have been extensively investigated to accelerate partial differential equations (PDE) solvers using data-driven surrogates. However, most data-driven surrogates require an extremely large amount of training data. In this paper, we propose the Convolutional Hierarchical Deep Learning Neural Network-Tensor Decomposition (C-HiDeNN-TD) method, which can directly obtain surrogate models by solving large-scale space-time PDE without generating any offline training data. We compare the performance of the proposed method against classical numerical methods for extremely large-scale systems.

cs.CE↗

arXiv2024

Multi-Patch Isogeometric Convolution Hierarchical Deep-learning Neural Network

A seamless integration of neural networks with Isogeometric Analysis (IGA) was first introduced in [1] under the name of Hierarchical Deep-learning Neural Network (HiDeNN) and has systematically evolved into Isogeometric Convolution HiDeNN (in short, C-IGA) [2]. C-IGA achieves higher order approximations without increasing the degree of freedom. Due to the Kronecker delta property of C-IGA shape functions, one can refine the mesh in the physical domain like standard finite element method (FEM) while maintaining the exact geometrical mapping of IGA. In this article, C-IGA theory is generalized for multi-CAD-patch systems with a mathematical investigation of the compatibility conditions at patch interfaces and convergence of error estimates. Two compatibility conditions (nodal compatibility and G^0 (i.e., global C^0) compatibility) are presented and validated through numerical examples.

math.NA↗

arXiv2024

Interpolating neural network: A novel unification of machine learning and interpolation theory

Artificial intelligence (AI) has revolutionized software development, shifting from task-specific codes (Software 1.0) to neural network-based approaches (Software 2.0). However, applying this transition in engineering software presents challenges, including low surrogate model accuracy, the curse of dimensionality in inverse design, and rising complexity in physical simulations. We introduce an interpolating neural network (INN), grounded in interpolation theory and tensor decomposition, to realize Engineering Software 2.0 by advancing data training, partial differential equation solving, and parameter calibration. INN offers orders of magnitude fewer trainable/solvable parameters for comparable model accuracy than traditional multi-layer perceptron (MLP) or physics-informed neural networks (PINN). Demonstrated in metal additive manufacturing, INN rapidly constructs an accurate surrogate model of Laser Powder Bed Fusion (L-PBF) heat transfer simulation, achieving sub-10-micrometer resolution for a 10 mm path in under 15 minutes on a single GPU. This makes a transformative step forward across all domains essential to engineering software.

cs.LG↗

arXiv2022

JAX-FEM: A differentiable GPU-accelerated 3D finite element solver for automatic inverse design and mechanistic data science

This paper introduces JAX-FEM, an open-source differentiable finite element method (FEM) library. Constructed on top of Google JAX, a rising machine learning library focusing on high-performance numerical computing, JAX-FEM is implemented with pure Python while scalable to efficiently solve problems with moderate to large sizes. For example, in a 3D tensile loading problem with 7.7 million degrees of freedom, JAX-FEM with GPU achieves around 10$\times$ acceleration compared to a commercial FEM code depending on platform. Beyond efficiently solving forward problems, JAX-FEM employs the automatic differentiation technique so that inverse problems are solved in a fully automatic manner without the need to manually derive sensitivities. Examples of 3D topology optimization of nonlinear materials are shown to achieve optimal compliance. Finally, JAX-FEM is an integrated platform for machine learning-aided computational mechanics. We show an example of data-driven multi-scale computations of a composite material where JAX-FEM provides an all-in-one solution from microscopic data generation and model training to macroscopic FE computations. The source code of the library and these examples are shared with the community to facilitate computational mechanics research.

cs.MS↗

arXiv2022

Deep Learning Discrete Calculus (DLDC): A Family of Discrete Numerical Methods by Universal Approximation for STEM Education to Frontier Research

The article proposes formulating and codifying a set of applied numerical methods, coined as Deep Learning Discrete Calculus (DLDC), that uses the knowledge from discrete numerical methods to interpret the deep learning algorithms through the lens of applied mathematics. The DLDC methods aim to leverage the flexibility and ever increasing resources of deep learning and rich literature of numerical analysis to formulate a general class of numerical method that can directly use data with uncertainty to predict the behavior of an unknown system as well as elevate the speed and accuracy of numerical solution of the governing equations for known systems. The article is structured in two major sections. In the first section, the building blocks of the DLDC methods are presented and deep learning structures analogous to traditional numerical methods such as finite difference and finite element methods are constructed with a view to incorporate these techniques in Science, Technology, Engineering, Mathematics (STEM) syllabus for K-12 students. The second section builds upon the building blocks of the previous discussion,and proposes new solution schemes for differential and integral equations pertinent to multiscale mechanics. Each section is accompanied with mathematical formulation of the numerical methods, analogous DLDC formulation, and suitable examples.

math.NA↗

arXiv2019

Jet modification with medium recoil in quark-gluon plasma

Jet energy transported to quark-gluon plasma during jet-medium interaction excites the QGP medium and creates energetic thermal partons -- recoil particles or recoils. Modification of the jet structure in heavy ion collisions is studied using \textsc{martini}, in which recoil simulation is enabled. In large systems such as central Pb-Pb collisions, the recoil effect is expected to be critical due to strong jet-medium interaction. We show the results of the jet mass function and jet shape function are improved when the recoil particles are included in the reconstructed jets. We conclude that the energy carried by the recoil particles are regarded as a part of reconstructed jets and are necessary in studying jet modification in heavy ion collisions.

nucl-th↗

arXiv2017

Multistage Monte-Carlo simulation of jet modification in a static medium

The modification of hard jets in an extended static medium held at a fixed temperature is studied using three different Monte-Carlo event generators (LBT, MATTER, MARTINI). Each event generator contains a different set of assumptions regarding the energy and virtuality of the partons within a jet versus the energy scale of the medium, and hence, applies to a different epoch in the space-time history of the jet evolution. For the first time, modeling is developed where a jet may sequentially transition from one generator to the next, on a parton-by-parton level, providing a detailed simulation of the space-time evolution of medium modified jets over a much broader dynamic range than has been attempted previously in a single calculation. Comparisons are carried out for different observables sensitive to jet quenching, including the parton fragmentation function and the azimuthal distribution of jet energy around the jet axis. The effect of varying the boundary between different generators is studied and a theoretically motivated criterion for the location of this boundary is proposed. The importance of such an approach with coupled generators to the modeling of jet quenching is discussed.

nucl-th↗

arXiv2016

Rapidity-dependent jet energy loss in small systems with finite-size effects and running coupling

Longitudinal dynamics of particle production and rapidity-dependent jet energy loss are investigated in small and asymmetric colliding systems. We utilize an improved version of \textsc{martini} in which two improvements are implemented to calculate the effect of the strongly coupled QGP droplet on jet energy loss. We show that those realistic prescriptions improve the results of nuclear modification factor calculations. We also observe visible energy loss of jets in a thermal background of high-multiplicity p-Pb collisions, and a clear correlation between the energy loss and elliptic flow coefficients for energetic particles. We conclude that systematic measurements of jet quenching in central collisions of small systems can support the formation of the QGP droplet.

nucl-th↗

arXiv2016

Direct photon production and jet energy-loss in small systems

Two types of penetrating probes, direct photon and QCD jets, are investigated in the background of a small and rapidly expanding droplet of quark-gluon plasma. The additional thermal electromagnetic radiation results in a $\sim$50\% enhancement of the direct photons. In high multiplicity p+Pb collisions, jets can lose a sizeable fraction of their initial energy, leading to a charged hadron $R_\mathrm{pA}$ of $\sim$0.8 at a transverse momentum around 10\,GeV. Those two proposed measurements can help understand the apparent collective behaviour observed in small collision systems.

hep-ph↗