arXiv subjects

Sungwoo Kang

Publications and source records attributed to Sungwoo Kang.

At least 19 recordsLinked to original sources

arXiv2026

The Limits of Complexity: Why Feature Engineering Beats Deep Learning in Investor Flow Prediction

The application of machine learning to financial prediction has accelerated dramatically, yet the conditions under which complex models outperform simple alternatives remain poorly understood. This paper investigates whether advanced signal processing and deep learning techniques can extract predictive value from investor order flows beyond what simple feature engineering achieves. Using a comprehensive dataset of 2.79 million observations spanning 2,439 Korean equities from 2020--2024, we apply three methodologies: \textit{Independent Component Analysis} (ICA) to recover latent market drivers, \textit{Wavelet Coherence} analysis to characterize multi-scale correlation structure, and \textit{Long Short-Term Memory} (LSTM) networks with attention mechanisms for non-linear prediction. Our results reveal a striking finding: a parsimonious linear model using market capitalization-normalized flows (``Matched Filter'' preprocessing) achieves a Sharpe ratio of 1.30 and cumulative return of 272.6\%, while the full ICA-Wavelet-LSTM pipeline generates a Sharpe ratio of only 0.07 with a cumulative return of $-5.1\%$. The raw LSTM model collapsed to predicting the unconditional mean, achieving a hit rate of 47.5\% -- worse than random. We conclude that in low signal-to-noise financial environments, domain-specific feature engineering yields substantially higher marginal returns than algorithmic complexity. These findings establish important boundary conditions for the application of deep learning to financial prediction.

q-fin.CP↗

arXiv2026

Forward-Facing Near-Infrared Adds Little to Colour for Farm-Machinery Traversability: A Site-Disjoint Evaluation of Sensor-Dependent Spatial Leakage

Near-infrared (NIR) imaging does not consistently outperform standard color cameras for daytime agricultural traversability once spatial data leakage is eliminated. Prior benchmarks suggesting an NIR advantage used sequence-level splits that permitted spatially autocorrelated imagery into test sets, artificially inflating NIR performance, especially on difficult paddy-boundary segmentation. When evaluated across strictly held-out recording sites using the AI Hub autonomous driving corpus, none of the four tested configurations (color, NIR, a luminance control, or their fusion) reliably surpasses standard color across any of the five traversability classes. Furthermore, apparent performance gains observed at a single held-out site completely fail to replicate when the held-out site is rotated, demonstrating that the observed benefits were site-specific artifacts rather than generalizable improvements. Correcting for spatial leakage erases the apparent lead of NIR entirely rather than uniformly degrading performance across all sensor types. Consequently, adding a forward-facing NIR camera to a daylight agricultural vehicle sensor suite remains unproven, and evaluations compromised by location leakage risk distorting sensor rankings and misleading procurement decisions.

cs.CV↗

arXiv2026

Region-Specific Calibration Achieves Excellent Inter-Device Reliability for Smartphone Dermatology: A Multi-Device Benchmark on Korean Facial Skin

Background: Smartphone-based dermatology requires inter-device colorimetric reliability that holds across calibration regimes, yet quantitative multi-device benchmarks remain scarce. Materials and Methods: We analyzed matched facial images from 965 Korean subjects captured by a digital single-lens reflex (DSLR) camera, a consumer tablet, and a consumer smartphone, and evaluated two calibration methods against the DSLR reference. The methods are standard global linear Color Correction Matrix (CCM) normalization and region-specific CCM trained per anatomical region, both applied in Commission Internationale de l'Eclairage Lab* (CIELAB) space. Results: Linear CCM reduced inter-device color differences by 61-74% and placed both Melanin Index (intraclass correlation coefficient [ICC] = 0.80) and Individual Typology Angle (ITA, ICC = 0.78) in the good reliability band. Region-specific CCM raised both indices into the excellent reliability band (MI ICC = 0.95, ITA ICC = 0.93), with anatomical region exceeding the source device as the largest pre-calibration variance contributor (analysis-of-variance $η^2 = 0.18$ versus 0.12). Conclusion: Consumer-device skin colorimetry therefore achieves clinically useful inter-device reliability using standard calibration, with region-aware calibration the largest remaining source of improvement.

eess.IV↗

arXiv2026

Asymmetry-Aware Routing for Industrial Multimodal Monitoring: A Diagnostic Framework

Multimodal fusion is the default approach for combining heterogeneous sensor streams in industrial monitoring, yet no systematic method exists for determining \textit{when fusion degrades rather than improves} detection performance. We present an \textbf{Asymmetry-Aware Routing Framework} -- a three-step diagnostic procedure (unimodal performance gap, gate weight attribution, modality corruption testing) with formal decision criteria -- that routes multimodal systems toward the appropriate fusion strategy before deployment. We validate the framework on three datasets spanning two routing outcomes: (1)~the OHT/AGV industrial dataset (thermal + sensors, 13{,}121 samples), where the framework correctly identifies severe asymmetry (gap ratio 3.1$\times$) and recommends \textsc{cascade}; (2)~a chain conveyor fault detection scenario (audio + vibration), where moderate asymmetry leads to a \textsc{fuse} recommendation with positive fusion benefit; and (3)~the CWRU bearing dataset, providing controlled validation in both directions. Threshold sensitivity analysis across all three datasets shows that the framework's recommendations are robust to threshold perturbation, with correct routing maintained over a wide parameter plateau. Comparison against simpler diagnostics (gap ratio alone) reveals that Step~1 alone is ambiguous for moderate-asymmetry cases, demonstrating the necessity of the full protocol for reliable routing decisions.

eess.SY↗

arXiv2026

Learning Structural Hardness for Combinatorial Auctions: Instance-Dependent Algorithm Selection via Graph Neural Networks

The Winner Determination Problem (WDP) in combinatorial auctions is NP-hard, and no existing method reliably predicts which instances will defeat fast greedy heuristics. The ML-for-combinatorial-optimization community has focused on learning to \emph{replace} solvers, yet recent evidence shows that graph neural networks (GNNs) rarely outperform well-tuned classical methods on standard benchmarks. We pursue a different objective: learning to predict \emph{when} a given instance is hard for greedy allocation, enabling instance-dependent algorithm selection. We design a 20-dimensional structural feature vector and train a lightweight MLP hardness classifier that predicts the greedy optimality gap with mean absolute error 0.033, Pearson correlation 0.937, and binary classification accuracy 94.7\% across three random seeds. For instances identified as hard -- those exhibiting ``whale-fish'' trap structure where greedy provably fails -- we deploy a heterogeneous GNN specialist that achieves ${\approx}0\%$ optimality gap on all six adversarial configurations tested (vs.\ 3.75--59.24\% for greedy). A hybrid allocator combining the hardness classifier with GNN and greedy solvers achieves 0.51\% overall gap on mixed distributions. Our honest evaluation on CATS benchmarks confirms that GNNs do not outperform Gurobi (0.45--0.71 vs.\ 0.20 gap), motivating the algorithm selection framing. Learning \emph{when} to deploy expensive solvers is more tractable than learning to replace them.

cs.LG↗

arXiv2026

Diagnosing Korean-Language LLM Political Bias via Census-Grounded Agent Simulation

Large language models (LLMs) exhibit systematic political biases in voter simulations, but their underlying mechanisms and cross-lingual generalizations remain poorly understood. We introduce Dynamo-K, a census-grounded simulation framework evaluating Korean-language LLM political behavior across four models on six Korean elections (2017-2025). Using this framework, we identify three systematic failure modes: (1) progressive bias in moderate agents, where explicit mitigation reduces Mean Absolute Error (MAE) by 5.2 times; (2) model-dependent third-party salience collapse, distinguishing between salience failure and decision bias; and (3) regional polarization collapse, where models bidirectionally under-predict historical party strongholds. To address these failures, we demonstrate that scenario reframing recovers 62% of 2017 MAE by restoring third-party visibility. Furthermore, we introduce a learned reweighting adapter that successfully calibrates opposing-valence models without relying on candidate names at train or test time. Validating our diagnostic framework, Dynamo-K accurately predicts 3/3 presidential winners - including a 2.1%p MAE on the highly contested 0.73%p-margin 2022 race - and correctly identifies the dominant party in a held-out local election. The pipeline is open-source and provides a scalable, cost-effective method for diagnosing LLM political behavior.

cs.CY↗

arXiv2026

STAG-CN: Spatio-Temporal Apiary Graph Convolutional Network for Disease Onset Prediction in Beehive Sensor Networks

Honey bee colony losses threaten global pollination services, yet current monitoring systems treat each hive as an isolated unit, ignoring the spatial pathways through which diseases spread across apiaries. This paper introduces the Spatio-Temporal Apiary Graph Convolutional Network (STAG-CN), a graph neural network that models inter-hive relationships for disease onset prediction. STAG-CN operates on a dual adjacency graph combining physical co-location and climatic sensor correlation among hive sessions, and processes multivariate IoT sensor streams through a temporal--spatial--temporal sandwich architecture built on causal dilated convolutions and Chebyshev spectral graph convolutions. Evaluated on the Korean AI Hub apiculture dataset (dataset \#71488) with expanding-window temporal cross-validation, STAG-CN achieves an F1 score of 0.607 at a three-day forecast horizon. An ablation study reveals that the climatic adjacency matrix alone matches full-model performance (F1\,=\,0.607), while the physical adjacency alone yields F1\,=\,0.274, indicating that shared environmental response patterns carry stronger predictive signal than spatial proximity for disease onset. These results establish a proof-of-concept for graph-based biosecurity monitoring in precision apiculture, demonstrating that inter-hive sensor correlations encode disease-relevant information invisible to single-hive approaches.

cs.LG↗

arXiv2026

Coarse-grained graph architectures for all-atom force predictions

We introduce a machine-learning framework termed coarse-grained all-atom force field (CGAA-FF), which incorporates coarse-grained message passing within an all-atom force field using equivariant nature of graph models. The CGAA-FF model employs grain embedding to encode atomistic coordinates into nodes representing grains rather than individual atoms, enabling predictions of both grain-level energies and atom-level forces. Tested on EC/EMC organic electrolytes and RDX crystalline and disordered phases, CGAA-FF achieves 0.201 and 0.253 eV A-1, respectively, while providing about 10-fold and 5-fold higher computational speed and memory efficiency, respectively, than conventional MLIPs. Since this CGAA framework can be integrated into any equivariant architecture, we believe this work opens the door to efficient all-atom simulations of soft-matter systems.

cond-mat.mtrl-sci↗

arXiv2026

Information Propagation Across Investor Types: Transfer Entropy Networks in the Korean Equity Market

Whether heterogeneous investor flows transmit private information across stocks or merely reflect coordinated responses to public signals remains an open question in market microstructure. We construct Transfer Entropy (TE) networks from investor-type flows -- foreign, institutional, and individual -- for \numNStocks{} Korean equities over \numNDates{} trading days (January 2020 to February 2025), and evaluate their economic content through interaction information (II), conditional TE, mutual information (MI), Kelly criterion bounds, and Fama-MacBeth regressions. Three findings emerge. First, TE networks are sparse and structurally heterogeneous: foreign investors maintain few but strong links (\numEdgesFor{} edges, mean TE = \numMeanTEFor{}), while individual investors form many but weak links (\numEdgesInd{} edges, mean TE = \numMeanTEInd{}). Second, cross-investor information is redundant rather than synergistic, no investor type directionally dominates another, and MI between signals and returns is zero at the daily horizon. Third, network centrality adds negligible alpha in cross-sectional regressions, with only one of six signal-centrality interactions reaching marginal significance. These results indicate that the observed propagation structure captures shared information processing rather than private signal cascades, consistent with daily-frequency market efficiency.

q-fin.ST↗

arXiv2026

When the Rules Change: Adaptive Signal Extraction via Kalman Filtering and Markov-Switching Regimes

Most empirical microstructure research assumes that order flow--return parameters are constant, yet these relationships shift substantially across market regimes. Combining adaptive Kalman filtering, Markov-switching regime identification, and asymmetric response estimation, we characterize regime-dependent investor behavior in the Korean stock market during 2020--2024 using daily transaction data disaggregated by investor type. Three principal findings emerge: foreign investor predictive power increases several-fold during crisis periods relative to bull markets; individual investors chase momentum asymmetrically, reacting far more strongly to positive than to negative shocks; and independent information-theoretic validation corroborates both patterns. Rigorous out-of-sample testing reveals that these in-sample regularities do not generalize reliably, underscoring the need for proper validation methodology in microstructure research.

q-fin.CP↗

arXiv2026

The Physics of Price Discovery: Deconvolving Information, Volatility, and the Critical Breakdown of Signal during Retail Herding

How information transmits through prices -- and why this transmission breaks down -- remains poorly understood. We combine regularized deconvolution with Hawkes process analysis to study the impulse response structure of investor flows in the Korean equity market (January 2020 -- February 2025). Three findings emerge: foreign and institutional flows drive permanent price discovery while individual flows provide contrarian liquidity; individual investor surges are predominantly panic-driven and exhibit near-explosive self-excitation; and during herding episodes, institutional price impact deteriorates sharply in small-cap stocks while large-cap stocks maintain resilience. These results reframe market efficiency as a state variable -- conditional on both herding intensity and firm size -- rather than a structural constant.

q-fin.ST↗

arXiv2026

Optimal Signal Extraction from Order Flow: A Matched Filter Perspective on Normalization and Market Microstructure

We establish a general matched filter principle for order flow normalization: optimal normalization must match the scaling behaviour of the signal-generating process. For capacity-constrained institutional investors, market capitalization normalization ($S^{MC}$) is the matched filter; for volume-targeting traders (e.g., VWAP/TWAP algorithms), trading value normalization ($S^{TV}$) is optimal. Monte Carlo simulations confirm this principle works bidirectionally, with matched filters achieving up to $1.99\times$ higher signal correlation. Empirical validation using 2.7 million stock-day observations from the Korean market (2020--2024) reveals symmetric normalization dominance across investor types: domestic institutional flows predict next-day returns significantly under $S^{MC}$ ($t = 9.65$), while foreign flows exhibit stronger predictability under $S^{TV}$ ($t = 16.35$) -- with no sign reversal at longer horizons, indicating durable private information rather than temporary price impact. These findings motivate the ``Informed Executor'' hypothesis: sophisticated foreign investors possess genuine private information but employ volume-targeting algorithms for stealth execution -- volume-scaling reflects execution methodology, not absence of information. Information-theoretic validation using KL divergence independently corroborates these results. The matched filter principle generalises to any market where signal scaling varies across trader types, with implications for trading algorithms, factor construction, and market microstructure methodology.

q-fin.CP↗

arXiv2026

Are diffusion models ready for materials discovery in unexplored chemical space?

While diffusion models are attracting increasing attention for the design of novel materials, their ability to generate low-energy structures in unexplored chemical spaces has not been systematically assessed. Here, we evaluate the performance of two diffusion models, MatterGen and DiffCSP, against three databases: a ternary oxide set (constructed by a genetic algorithm), a ternary nitride set (constructed by template informatics), and the GNoME database (constructed by a combination of both). We find that diffusion models generally perform stably in well-sampled chemical spaces (oxides and nitrides), but are less effective in uncommon ones (GNoME), which contains many compositions involving rare-earth elements and unconventional stoichiometry. Finally, we assess their size-extrapolation capability and observe a significant drop in performance when the number of atoms exceeds the trained range. This is attributed to the limitations imposed by periodic boundary conditions, which we refer to as the curse of periodicity. This study paves the way for future developments in materials design by highlighting both the strength and the limitations of diffusion models.

cond-mat.mtrl-sci↗

arXiv2026

Multi-Task Learning for Metal Alloy Property Prediction: An Empirical Study of Negative Transfer and Mitigation Strategies

Multi-task learning (MTL) in materials science relies on the assumption that physically related properties share learnable representations. We challenge this assumption using a 54,028-sample metal alloy dataset exhibiting extreme task-level imbalance. Our results reveal a striking dichotomy: MTL significantly degrades regression performance for resistivity and hardness but improves classification recall for amorphous-forming ability. We trace this divergence to mismatched functional forms--such as resistivity's polynomial dependence versus hardness's complex interactions--which cause severe gradient misalignment during optimization. Evaluating Deep Imbalanced Regression techniques, we find that projecting conflicting gradients (PCGrad) recovers single-task performance, while combining label distribution smoothing with gradient normalization achieves the best overall balance. Consequently, we propose a strategic framework: utilize independent models for high-precision characterization, but employ MTL for high-throughput screening where recall is paramount. These findings support a "materials property clustering" hypothesis, suggesting that distinct physical mechanisms require specialized optimization strategies to overcome negative transfer.

cs.LG↗

arXiv2025

Sources and Nonlinearity of High Volume Return Premium: An Empirical Study on the Differential Effects of Investor Identity versus Trading Intensity (2020-2024)

Chae and Kang (2019, \textit{Pacific-Basin Finance Journal}) documented a puzzling Low Volume Return Premium (LVRP) in Korea -- contradicting global High Volume Return Premium (HVRP) evidence. We resolve this puzzle. Using Korean market data (2020-2024), we demonstrate that HVRP exists in Korea but is masked by (1) pooling heterogeneous investor types and (2) using inappropriate intensity normalization. When institutional buying intensity is normalized by market capitalization rather than trading value, a perfect monotonic relationship emerges: highest-conviction institutional buying (Q4) generates +\institutionLedQFourDayPlusFiftyCAR\ cumulative abnormal returns over 50 days, while lowest-intensity trades (Q1) yield modest returns (+\institutionLedQOneDayPlusFiftyCAR). Retail investors exhibit a flat pattern -- their trading generates near-zero returns regardless of conviction level -- confirming the pure noise trader hypothesis. During the Donghak Ant Movement (2020-2021), however, coordinated retail investors temporarily transformed from noise traders to liquidity providers, generating returns comparable to institutional trading. Our findings reconcile conflicting international evidence and demonstrate that detecting informed trading signals requires investor-type decomposition, nonlinear quartile analysis, and conviction-based (market cap) rather than participation-based (trading value) measurement.

q-fin.TR↗

arXiv2025

Modern Deep Learning Approaches for Cricket Shot Classification: A Comprehensive Baseline Study

Cricket shot classification from video sequences remains a challenging problem in sports video analysis, requiring effective modeling of both spatial and temporal features. This paper presents the first comprehensive baseline study comparing seven different deep learning approaches across four distinct research paradigms for cricket shot classification. We implement and systematically evaluate traditional CNN-LSTM architectures, attention-based models, vision transformers, transfer learning approaches, and modern EfficientNet-GRU combinations on a unified benchmark. A critical finding of our study is the significant performance gap between claims in academic literature and practical implementation results. While previous papers reported accuracies of 96\% (Balaji LRCN), 99.2\% (IJERCSE), and 93\% (Sensors), our standardized re-implementations achieve 46.0\%, 55.6\%, and 57.7\% respectively. Our modern SOTA approach, combining EfficientNet-B0 with a GRU-based temporal model, achieves 92.25\% accuracy, demonstrating that substantial improvements are possible with modern architectures and systematic optimization. All implementations follow modern MLOps practices with PyTorch Lightning, providing a reproducible research platform that exposes the critical importance of standardized evaluation protocols in sports video analysis research.

cs.CV↗

arXiv2025

Discovery of oxide Li-conducting electrolytes in uncharted chemical space via topology-constrained crystal structure prediction

Oxide Li-conducting solid-state electrolytes (SSEs) offer excellent chemical and thermal stability but typically exhibit lower ionic conductivity than sulfides and chlorides. This motivates the search for new oxide materials with enhanced conductivity. Crystal structure prediction is a powerful approach for identifying such candidates. However, the structural complexity of oxide SSEs, often involving unit cells with more than 100 atoms, presents significant challenges for conventional methods. In this study, we introduce TOPIC, a structure prediction algorithm that reduces configurational complexity by enforcing corner-sharing (CS) bond topology constraints. We demonstrate that TOPIC successfully reproduces the ground-state and metastable structures of known oxide SSEs, including LiTa$_2$PO$_8$ and Li$_7$La$_3$Zr$_2$O$_{12}$, which contain up to about 200 atoms per unit cell. By combining this approach with a pretrained machine-learning interatomic potential, we systematically screen quaternary oxide compositions and identify 92 promising candidates with CS frameworks. In particular, Li$_4$Hf$_2$Si$_3$O$_{12}$, which corresponds to the ground state at its composition, exhibits an ionic conductivity of 14 mS cm$^{-1}$, a hull energy of 21 meV atom$^{-1}$, and a band gap of 6.5 eV. Through our investigation, we identify the Li ratio as one of the key factors determining the stability of CS structures. Overall, our approach provides a practical and scalable pathway for discovering high-performance oxide solid electrolytes in previously unexplored chemical spaces.

cond-mat.mtrl-sci↗

arXiv2025

An efficient forgetting-aware fine-tuning framework for pretrained universal machine-learning interatomic potentials

Pretrained universal machine-learning interatomic potentials (MLIPs) have revolutionized computational materials science by enabling rapid atomistic simulations as efficient alternatives to ab initio methods. Fine-tuning pretrained MLIPs offers a practical approach to improving accuracy for materials and properties where predictive performance is insufficient. However, this approach often induces catastrophic forgetting, undermining the generalizability that is a key advantage of pretrained MLIPs. Herein, we propose reEWC, an advanced fine-tuning strategy that integrates Experience Replay and Elastic Weight Consolidation (EWC) to effectively balance forgetting prevention with fine-tuning efficiency. Using Li$_6$PS$_5$Cl (LPSC), a sulfide-based Li solid-state electrolyte, as a fine-tuning target, we show that reEWC significantly improves the accuracy of a pretrained MLIP, resolving well-known issues of potential energy surface softening and overestimated Li diffusivities. Moreover, reEWC preserves the generalizability of the pretrained MLIP and enables knowledge transfer to chemically distinct systems, including other sulfide, oxide, nitride, and halide electrolytes. Compared to Experience Replay and EWC used individually, reEWC delivers clear synergistic benefits, mitigating their respective limitations while maintaining computational efficiency. These results establish reEWC as a robust and effective solution for continual learning in MLIPs, enabling universal models that can advance materials research through large-scale, high-throughput simulations across diverse chemistries.

cond-mat.mtrl-sci↗