arXiv subjects

Yongjoo Kim

Publications and source records attributed to Yongjoo Kim.

4 recordsLinked to original sources

arXiv2026

SAGE: Semantic Attribute Graphs for Multi-Entity Visual Retrieval

Dense document images often contain many fine-grained visual and textual entities whose relevance depends on a user query. Standard vision-language retrievers encode cropped regions with a single vector, which can mix distinct entity signals and obscure the evidence needed for fine-grained retrieval. We call this failure mode Semantic Dilution and quantitatively show that it degrades entity-level retrieval as a function of entity density. To mitigate it, we propose SAGE, a training-free framework that parses semantic entities from dense document images, represents them as hierarchical graph nodes with multi-vector embeddings, and retrieves query-relevant evidence through iterative entity-level subgraph matching. We also introduce DEAR, a dataset of 1,055 query--image pairs sourced from product detail pages, where each query requires retrieving and comparing multiple fine-grained entities from visually dense inputs across four question types of increasing complexity. Experiments show that SAGE substantially reduces semantic dilution and outperforms patch-level and OCR-based retrieval baselines on DEAR, achieving a Recall@3 of 0.849 and a generation score of 2.746 on multi-entity visual comparison queries. Our code is available at https://github.com/All4Nothing/SAGE.

cs.IR↗

arXiv2026

Broadband Second Harmonic Generation using Fixed-Period Periodically Poled Lithium Niobate

Periodically poled lithium niobate (PPLN) is a widely used nonlinear optical device for second harmonic generation (SHG). Despite its wide adoption in commercial systems, its bandwidth for SHG is fundamentally limited by the quasi-phase matching condition. This can be overcome by aperiodic or chirped PPLN structures; however, such devices are typically custom-fabricated and not readily available commercially. In this study, we investigate an alternative approach to achieving broadband SHG by using a standard PPLN crystal containing multiple fixed poling periods. Broadband operation is realized by angle tuning the crystal relative to input beam in free-space. The effect of angle tuning is examined over a range of incident angles, and an 1.6x enhancement in SHG bandwidth is demonstrated. These results suggest a practical and efficient strategy for broadband SHG using standard PPLN devices.

physics.optics↗

arXiv2024

Network Representation Learning for Biophysical Neural Network Analysis

The analysis of biophysical neural networks (BNNs) has been a longstanding focus in computational neuroscience. A central yet unresolved challenge in BNN analysis lies in deciphering the correlations between neuronal and synaptic dynamics, their connectivity patterns, and learning process. To address this, we introduce a novel BNN analysis framework grounded in network representation learning (NRL), which leverages attention scores to uncover intricate correlations between network components and their features. Our framework integrates a new computational graph (CG)-based BNN representation, a bio-inspired graph attention network (BGAN) that enables multiscale correlation analysis across BNN representations, and an extensive BNN dataset. The CG-based representation captures key computational features, information flow, and structural relationships underlying neuronal and synaptic dynamics, while BGAN reflects the compositional structure of neurons, including dendrites, somas, and axons, as well as bidirectional information flows between BNN components. The dataset comprises publicly available models from ModelDB, reconstructed using the Python and standardized in NeuroML format, and is augmented with data derived from canonical neuron and synapse models. To our knowledge, this study is the first to apply an NRL-based approach to the full spectrum of BNNs and their analysis.

cs.LG↗

arXiv2023

Needle guidance with Doppler-tracked polarization-sensitive optical coherence tomography

We demonstrate that a simple, unscanned polarization-sensitive optical coherence tomography needle probe can be used to perform layer identification in biological tissues. Broadband light from a laser centered at 1310 nm was sent through a fiber that was embedded into a needle, and analysis of the polarization state of the returning light after interference coupled with Doppler-based tracking allowed the calculation of phase retardation and optic axis orientation at each needle location. Proof-of-concept phase retardation mapping was shown in Atlantic salmon tissue, while axis orientation mapping was demonstrated in white shrimp tissue. The needle probe was then tested on the ex vivo porcine spine, where mock epidural procedures were performed. Our imaging results demonstrate that unscanned, Doppler-tracked polarization-sensitive optical coherence tomography imaging successfully identified the skin, subcutaneous tissue, and ligament layers, before successfully reaching the target of the epidural space. The addition of polarization-sensitive imaging into the bore of a needle probe therefore allows layer identification at deeper locations in the tissue.

physics.med-ph↗