arXiv subjects

Jungmin Kim

Publications and source records attributed to Jungmin Kim.

11 recordsLinked to original sources

arXiv2026

Cross-Temperature Defect Identification in Atomistic Simulations via Multi-Level Domain Alignment

Identifying atomic defects at elevated temperature is difficult because thermal fluctuations blur the local symmetry that both geometric heuristics and supervised classifiers rely on: trustworthy labels exist in low-temperature reference configurations, while the high-temperature regime where robust analysis matters most is effectively unlabeled. We cast this as a cross-temperature domain-shift problem and align the two domains at three levels: an equivariant denoiser at the input level, cross-temperature contrastive learning at the representation level, and a morphology-aware regularizer that steers predictions toward the compact geometry of physical defect structures. Because no atom-wise truth exists at temperature, we further introduce a label-free evaluation suite that scores predicted defect structures along five spatial and physics-based axes, enabling model assessment and selection without high-temperature labels. Near the melting point, the framework identifies vacancies and self-interstitial atoms across face-centered-cubic, body-centered-cubic, and hexagonal-close-packed iron systems with every interstitial localized and zero false detections in every vacancy system against Wigner-Seitz ground truth, with no high-temperature labels used in training. It sustains this fidelity on a million-atom, 2.5 ns trajectory, resolving single vacancy hops and complete Frenkel-pair recombination, and captures grain-boundary phase transformations in aluminum bicrystals, distinguishing two nucleation modes. Multi-level domain alignment thus offers a practical, label-efficient route to temperature-robust structural analysis of large-scale molecular dynamics.

cond-mat.mtrl-sci↗

arXiv2026

Tailoring Defects in Photonic Time Crystals for Coherent Energy Control

Recent advances in time-varying photonics have revealed new degrees of freedom for manipulating optical states, arising from the distinctive nature of the temporal axis: causality and open-system dynamics. A representative example is photonic time crystals (PTCs) characterized by discrete time-translational symmetry, which exhibit space-analogous yet distinct phenomena, such as momentum gaps and amplifying-decaying Floquet-mode pairs. Although PTCs enable optical-energy amplification beyond conventional gain media, their application as programmable energy-functional devices remains challenging. Here, we propose a design framework for tailoring optical energy via defective PTCs. By optimizing defect permittivity and duration using analytic gradients of time transfer matrices, we realize prescribed coherent energy amplification and suppression. We show that a single defect enables continuous energy tailoring, while revealing an intrinsic asymmetry between amplification and suppression due to the inherently amplifying nature of the momentum gap. Extending the framework to coupled defects expands the design space and markedly improves suppression, establishing temporal-defect engineering as a route to programmable coherent energy control.

physics.optics↗

arXiv2026

Self-Abstraction Learning for Effective and Stable Training of Deep Neural Networks

Training large-scale deep neural networks effectively and stably is essential for applying deep learning across various fields. However, conventional methods, which rely on training a single large network, often encounter challenges such as gradient vanishing, overfitting and unstable learning. To overcome these limitations, we introduce Self-Abstraction Learning (SAL), a hierarchical framework. In SAL, networks are arranged by structural complexity, where the simplest topmost network is trained first and its hidden and output layers serve as guidance for the successively more complex networks below. This top-down sequential guidance effectively mitigates optimization issues, enabling stable training of deep architectures. Various experiments across MLP, CNN, and RNN architectures demonstrate that SAL consistently outperforms conventional methods, ensuring robust generalization even in data-scarce and complex network regimes.

cs.LG↗

arXiv2026

Quantum Nonlinearity for Optical Neural Computing

The rapid scaling of deep neural networks comes at the cost of unsustainable power consumption. While optical neural networks offer an alternative, their capabilities remain constrained by the lack of efficient optical nonlinearities. To address this, we propose an optical neural computing architecture by embedding quantum emitters in inverse-designed nanophotonic structures. Due to their saturability, quantum emitters exhibit exceptionally strong nonlinearity compared with conventional materials. Using physics-aware training, we numerically demonstrate that the proposed architecture can solve complex tasks, including nonlinear classification and reinforcement learning, within all-optical neural networks. To enable fair comparison across different platforms, we introduce a framework that quantitatively links nonlinearity to a network's expressive power. Analysis shows that our quantum activation operates at $\text{nW}/\mu\text{m}^2$ intensity, which is seven orders of magnitude below the nonlinearity threshold of conventional optical materials. Looking ahead to large language models, we estimate the nonlinearity-limited optical power, which scales sublinearly with model size. Our results indicate that quantum nanophotonics may provide a route toward sustainable AI inference.

physics.optics↗

arXiv2024

Photonic systolic array for all-optical matrix-matrix multiplication

Systolic arrays have proven to be highly efficient for parallelized matrix-matrix multiplication (MMM), utilizing synchronized, heartbeat-like data flows across an array of processing elements. While optical structures such as waveguide crossbar arrays and Mach-Zehnder interferometer-based meshes serve as photonic equivalents to the systolic arrays, the disparity between the two input matrices for multiplication -- one using optical signals and the other with system-defined parameters -- gives rise to a bottleneck in modern machine-learning tasks, such as evaluating attention scores in large language models. Here, we propose a photonic systolic array that performs MMM entirely with optical signals, utilizing homodyne detection at each array cell. Adjoint-based design of compact on-chip freeform optical modules enables precise control of light flow without bulky waveguide coupling schemes. The operation of a $4\times4$ photonic systolic array is numerically verified, achieving a theoretical computation density of $6.2~\mathrm{PMACs}/\mathrm{mm}^2/\mathrm{s}$. This design marks a significant step toward practical photonic computing hardware for modern AI workloads.

physics.optics↗

arXiv2024

Compute-first optical detection for noise-resilient visual perception

In the context of visual perception, the optical signal from a scene is transferred into the electronic domain by detectors in the form of image data, which are then processed for the extraction of visual information. In noisy and weak-signal environments such as thermal imaging for night vision applications, however, the performance of neural computing tasks faces a significant bottleneck due to the inherent degradation of data quality upon noisy detection. Here, we propose a concept of optical signal processing before detection to address this issue. We demonstrate that spatially redistributing optical signals through a properly designed linear transformer can enhance the detection noise resilience of visual perception tasks, as benchmarked with the MNIST classification. Our idea is supported by a quantitative analysis detailing the relationship between signal concentration and noise robustness, as well as its practical implementation in an incoherent imaging system. This compute-first detection scheme can pave the way for advancing infrared machine vision technologies widely used for industrial and defense applications.

physics.optics↗

arXiv2024

Towards Diverse Perspective Learning with Selection over Multiple Temporal Poolings

In Time Series Classification (TSC), temporal pooling methods that consider sequential information have been proposed. However, we found that each temporal pooling has a distinct mechanism, and can perform better or worse depending on time series data. We term this fixed pooling mechanism a single perspective of temporal poolings. In this paper, we propose a novel temporal pooling method with diverse perspective learning: Selection over Multiple Temporal Poolings (SoM-TP). SoM-TP dynamically selects the optimal temporal pooling among multiple methods for each data by attention. The dynamic pooling selection is motivated by the ensemble concept of Multiple Choice Learning (MCL), which selects the best among multiple outputs. The pooling selection by SoM-TP's attention enables a non-iterative pooling ensemble within a single classifier. Additionally, we define a perspective loss and Diverse Perspective Learning Network (DPLN). The loss works as a regularizer to reflect all the pooling perspectives from DPLN. Our perspective analysis using Layer-wise Relevance Propagation (LRP) reveals the limitation of a single perspective and ultimately demonstrates diverse perspective learning of SoM-TP. We also show that SoM-TP outperforms CNN models based on other temporal poolings and state-of-the-art models in TSC with extensive UCR/UEA repositories.

cs.LG↗

arXiv2022

Unidirectional scattering with spatial homogeneity using photonic time disorder

The temporal degree of freedom in photonics has been a recent research hotspot due to its analogy with spatial axes, causality, and open-system characteristics. In particular, the temporal analogues of photonic crystals have stimulated the design of momentum gaps and their extension to topological and non-Hermitian photonics. Although recent studies have also revealed the effect of broken discrete time-translational symmetry in view of the temporal analogy of spatial Anderson localization, the broad intermediate regime between time order and time uncorrelated disorder has not been examined. Here we investigate the inverse design of photonic time disorder to achieve optical functionalities in spatially homogeneous platforms. By developing the structure factor and order metric using causal Green's functions for the domain of time disorder, we demonstrate engineered time scatterer, which provides unidirectional scattering with controlled scattering amplitudes. We also reveal that the order-to-disorder transition in the time domain allows for the manipulation of scattering bandwidths, which inspires resonance-free temporal colour filtering. Our work will pave the way for advancing optical functionalities without spatial patterning.

physics.optics↗

arXiv2015

Little strings and T-duality

We study the 2d N = 4 gauge theory descriptions of little strings on type II NS5- branes. The IIB strings on N NS5-branes are described by the N = (4,4) gauge theories, whose Higgs branch CFTs on U(N) instanton moduli spaces are relevant. The IIA strings are described by N = (4,4) circular A_{N-1} quiver theories, whose Coulomb branch CFTs are relevant. We study new N = (0,4) quiver gauge theories for the IIA strings, which make it easier to study some infrared observables. In particular, we show that the elliptic genera of the IIA / IIB strings precisely map to each other by T-duality.

hep-th↗

arXiv2014

Super-Yang-Mills theories on S4 x R

We construct super-Yang-Mills theories on S4 x R, S4 x S1 and S4 x interval with the field content of maximal SYM, coupled to boundary degrees in the last case. These theories provide building blocks of the `5d uplifts' of gauge theories on S4, obtained by compactifying the 6d (2,0) theory. We pay special attention to the N=2* theory on S4. We also explain how to construct maximal SYM on S5 x R, and clarify when SYM theories can be put on S^n x R.

hep-th↗

arXiv2012

Vortices and 3 dimensional dualities

We study a supersymmetric partition function of topological vortices in 3d N=4,3 gauge theories on R^2 x S^1, and use it to explore Seiberg-like dualities with Fayet-Iliopoulos deformations. We provide a detailed support of these dualities and also clarify the roles of vortices. The N=4 partition function confirms the proposed Seiberg duality and suggests nontrivial extensions, presumably at novel IR fixed points with enhanced symmetries. The N=3 theories with nonzero Chern-Simons term also have non-topological vortices in the partially broken phases, which are essential for the Seiberg duality invariance of the spectrum. We use our partition function to confirm some properties of non-topological vortices via Seiberg duality in a simple case.

hep-th↗