arXiv subjects

Dongjin Seo

Publications and source records attributed to Dongjin Seo.

7 recordsLinked to original sources

arXiv2025

Generalizing Deep Surrogate Solvers for Broadband Electromagnetic Field Prediction at Unseen Wavelengths

Recently, electromagnetic surrogate solvers, trained on solutions of Maxwell's equations under specific simulation conditions, enabled fast inference of computationally expensive simulations. However, conventional electromagnetic surrogate solvers often consider only a narrow range of spectrum and fail when encountering even slight variations in simulation conditions. To address this limitation, we define spectral consistency as the property by which the spatial frequency structure of wavelength-dependent condition embeddings matches that of the target electromagnetic field patterns. In addition, we propose two complementary components: a refined wave prior, which is the condition embedding that satisfies spectral consistency, and Wave-Informed element-wise Multiplicative Encoding (WIME), which integrates these embeddings throughout the model while preserving spectral consistency. This framework enables accurate field prediction across the broadband spectrum, including untrained intermediate wavelengths. Our approach reduces the normalized mean squared error at untrained wavelengths by up to 71% compared to the state-of-the-art electromagnetic surrogate solver and achieves a speedup of over 42 times relative to conventional numerical simulations.

cs.LG↗

arXiv2025

Physics-guided and fabrication-aware inverse design of photonic devices using diffusion models

Designing free-form photonic devices is fundamentally challenging due to the vast number of possible geometries and the complex requirements of fabrication constraints. Traditional inverse-design approaches--whether driven by human intuition, global optimization, or adjoint-based gradient methods--often involve intricate binarization and filtering steps, while recent deep learning strategies demand prohibitively large numbers of simulations (10^5 to 10^6). To overcome these limitations, we present AdjointDiffusion, a physics-guided framework that integrates adjoint sensitivity gradients into the sampling process of diffusion models. AdjointDiffusion begins by training a diffusion network on a synthetic, fabrication-aware dataset of binary masks. During inference, we compute the adjoint gradient of a candidate structure and inject this physics-based guidance at each denoising step, steering the generative process toward high figure-of-merit (FoM) solutions without additional post-processing. We demonstrate our method on two canonical photonic design problems--a bent waveguide and a CMOS image sensor color router--and show that our method consistently outperforms state-of-the-art nonlinear optimizers (such as MMA and SLSQP) in both efficiency and manufacturability, while using orders of magnitude fewer simulations (approximately 2 x 10^2) than pure deep learning approaches (approximately 10^5 to 10^6). By eliminating complex binarization schedules and minimizing simulation overhead, AdjointDiffusion offers a streamlined, simulation-efficient, and fabrication-aware pipeline for next-generation photonic device design. Our open-source implementation is available at https://github.com/dongjin-seo2020/AdjointDiffusion.

physics.optics↗

arXiv2024

LPU: A Latency-Optimized and Highly Scalable Processor for Large Language Model Inference

The explosive arrival of OpenAI's ChatGPT has fueled the globalization of large language model (LLM), which consists of billions of pretrained parameters that embodies the aspects of syntax and semantics. HyperAccel introduces latency processing unit (LPU), a latency-optimized and highly scalable processor architecture for the acceleration of LLM inference. LPU perfectly balances the memory bandwidth and compute logic with streamlined dataflow to maximize performance and efficiency. LPU is equipped with expandable synchronization link (ESL) that hides data synchronization latency between multiple LPUs. HyperDex complements LPU as an intuitive software framework to run LLM applications. LPU achieves 1.25 ms/token and 20.9 ms/token for 1.3B and 66B model, respectively, which is 2.09x and 1.37x faster than the GPU. LPU, synthesized using Samsung 4nm process, has total area of 0.824 mm2 and power consumption of 284.31 mW. LPU-based servers achieve 1.33x and 1.32x energy efficiency over NVIDIA H100 and L4 servers, respectively.

cs.AR↗

arXiv2023

A3SA: Advanced Data Augmentation via Adjoint Sensitivity Analysis

Innovative machine learning techniques have facilitated the inverse design of photonic structures for numerous practical applications. Nevertheless, within these approaches, the quantity of data and the initial data distribution are paramount for the discovery of highly efficient photonic devices. These devices often require simulated data ranging from thousands to several hundred thousand data points. This issue has consistently posed a major hurdle in machine learning-based photonic design problems. Therefore, we propose a novel data augmentation algorithm grounded in the adjoint method, capable of generating more than 300 times the amount of original data while enhancing device efficiency. The adjoint method forecasts changes in the figure of merit (FoM) resulting from structural perturbations, requiring only two full-wave Maxwell simulations for this prediction. By leveraging the adjoint gradient values, we can augment and label several thousand new data points without any additional computations. Furthermore, the augmented data generated by the proposed algorithm displays significantly improved FoMs owing to the precise FoM change predictions enabled by the adjoint gradients. We apply this algorithm to a multi-layered metalens design problem and demonstrate that it consequently exhibits a 343-fold increase in data generation efficiency. After incorporating the proposed algorithm into a generative adversarial network (GAN), the optimized metalens exhibits a maximum focusing efficiency of 92.93%, comparable to the theoretical upper bound (93.80%).

physics.optics↗

arXiv2023

Physics-informed reinforcement learning for sample-efficient optimization of freeform nanophotonic devices

In the field of optics, precise control of light with arbitrary spatial resolution has long been a sought-after goal. Freeform nanophotonic devices are critical building blocks for achieving this goal, as they provide access to a design potential that could hardly be achieved by conventional fixed-shape devices. However, finding an optimal device structure in the vast combinatorial design space that scales exponentially with the number of freeform design parameters has been an enormous challenge. In this study, we propose physics-informed reinforcement learning (PIRL) as an optimization method for freeform nanophotonic devices, which combines the adjoint-based method with reinforcement learning to enhance the sample efficiency of the optimization algorithm and overcome the issue of local minima. To illustrate these advantages of PIRL over other conventional optimization algorithms, we design a family of one-dimensional metasurface beam deflectors using PIRL, obtaining more performant devices. We also explore the transfer learning capability of PIRL that further improves sample efficiency and demonstrate how the minimum feature size of the design can be enforced in PIRL through reward engineering. With its high sample efficiency, robustness, and ability to seamlessly incorporate practical device design constraints, our method offers a promising approach to highly combinatorial freeform device optimization in various physical domains.

physics.comp-ph↗

arXiv2013

Physical Principles for Scalable Neural Recording

Simultaneously measuring the activities of all neurons in a mammalian brain at millisecond resolution is a challenge beyond the limits of existing techniques in neuroscience. Entirely new approaches may be required, motivating an analysis of the fundamental physical constraints on the problem. We outline the physical principles governing brain activity mapping using optical, electrical,magnetic resonance, and molecular modalities of neural recording. Focusing on the mouse brain, we analyze the scalability of each method, concentrating on the limitations imposed by spatiotemporal resolution, energy dissipation, and volume displacement. We also study the physics of powering and communicating with microscale devices embedded in brain tissue.

q-bio.NC↗

arXiv2013

Neural Dust: An Ultrasonic, Low Power Solution for Chronic Brain-Machine Interfaces

A major hurdle in brain-machine interfaces (BMI) is the lack of an implantable neural interface system that remains viable for a lifetime. This paper explores the fundamental system design trade-offs and ultimate size, power, and bandwidth scaling limits of neural recording systems built from low-power CMOS circuitry coupled with ultrasonic power delivery and backscatter communication. In particular, we propose an ultra-miniature as well as extremely compliant system that enables massive scaling in the number of neural recordings from the brain while providing a path towards truly chronic BMI. These goals are achieved via two fundamental technology innovations: 1) thousands of 10 - 100 μm scale, free-floating, independent sensor nodes, or neural dust, that detect and report local extracellular electrophysiological data, and 2) a sub-cranial interrogator that establishes power and communication links with the neural dust.

q-bio.NC↗