Selected work

Representative Papers

Upper Bounds on the Generalization Error of Deep Learning Models via Local Robustness and Stability

Jun 15, 2026

Existing robustness-based generalization error bounds are often excessively loose or even vacuous in practice, failing to reflect the true generalization capability of deep models. This work proposes a locally adaptive approach to constructing generalization bounds by introducing notions of local robustness and stability. The input space is partitioned into subregions, and the robustness term is scaled according to the proportion of stable versus unstable samples within each region, yielding a tighter, data- and model-dependent upper bound. This method is the first to refine global robustness measures into localized forms, substantially mitigating the vacuity commonly observed in traditional bounds. Experiments on ImageNet demonstrate that the proposed bound remains non-vacuous across various robust deep networks and closely tracks their empirical generalization errors, significantly outperforming existing approaches.

0 citationsRead paper

Advances in Battery Energy Storage Management: Control and Economic Synergies

Feb 06, 2026

This study addresses the common disconnect in existing battery energy storage system research between economic dispatch and operational control, which often compromises the dual objectives of grid stability and revenue maximization. By systematically reviewing five key areas—ancillary service modeling, real-time power control, optimization-based scheduling algorithms, techno-economic analysis, and digital twin technologies—the work proposes a synergistic framework that deeply integrates grid dispatch economics with battery control systems. The research identifies critical gaps in cross-domain coordination and establishes a digital twin architecture for energy storage that simultaneously supports high profitability and high reliability. This framework provides both theoretical foundations and practical technical pathways for the development of next-generation intelligent energy storage systems.

0 citationsRead paper

A Physics Informed Neural Network For Deriving MHD State Vectors From Global Active Regions Observations

Dec 23, 2025

Accurate early prediction of solar active regions (ARs) requires physically consistent initialization of the tachocline—the sub-photospheric magnetohydrodynamic (MHD) dynamo layer—yet conventional inversion methods lack physical self-consistency. Method: We propose PINNBARDS, a novel physics-informed neural network (PINN) framework that embeds the shallow-water MHD equations as hard constraints to enable end-to-end reconstruction of dynamically evolving, energy- and momentum-conserving tachocline initial states from observed photospheric magnetic ribbon geometry (e.g., SDO/HMI data). Contribution/Results: Applied to 14 February 2024 observations, PINNBARDS successfully reconstructed antisymmetric active band structures with optimal toroidal field strength of 20–30 kG and latitudinal width ≈10°, in excellent agreement with low-order radial-mode excitation theory. This constitutes the first physically grounded, observationally constrained initialization scheme enabling week-scale forecasting of AR emergence and associated flare activity.

0 citationsRead paper

A Validation Strategy for Deep Learning Models: Evaluating and Enhancing Robustness

Sep 23, 2025

Deep learning models exhibit insufficient robustness against adversarial perturbations and common image corruptions, undermining their reliability in real-world deployment. To address this, we propose an active robustness verification strategy that leverages the training set itself: by performing local robustness analysis, our method automatically identifies “weakly robust” samples—serving as early, interpretable indicators of model vulnerability—and enables targeted robustness enhancement. Unlike conventional passive paradigms that rely solely on perturbed test sets for robustness evaluation, ours is the first to repurpose training data for robustness diagnostics. We integrate adversarial perturbation injection with diverse natural corruption tests. Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet demonstrate that our strategy significantly improves model robustness against both attacks and corruptions (average gain of +8.2%) while enhancing the sensitivity and interpretability of reliability assessment.

0 citationsRead paper

Scalable Strategies for Continual Learning with Replay

May 18, 2025

Existing replay-based continual learning methods suffer from poor scalability and high computational overhead, while failing to effectively integrate parameter-efficient fine-tuning techniques—such as Low-Rank Adaptation (LoRA) and model merging—into multi-task adaptation. To address this, we propose a novel replay framework that deeply couples staged knowledge consolidation with sequential model merging tailored for task-ordered streams. It is the first to systematically integrate LoRA, task arithmetic, and collaborative optimization within a unified replay paradigm. Our approach reduces required replay sample volume by up to 55%, significantly improving training efficiency and model scalability. Empirically, it outperforms individual baseline methods across diverse multi-task sequences, achieving organic synergy between replay mechanisms and parameter-efficient adaptation. This work establishes a new paradigm for lightweight, accumulative continual learning models.

0 citationsRead paper

Recent publications

Latest Papers

Upper Bounds on the Generalization Error of Deep Learning Models via Local Robustness and Stability

Jun 15, 2026

Existing robustness-based generalization error bounds are often excessively loose or even vacuous in practice, failing to reflect the true generalization capability of deep models. This work proposes a locally adaptive approach to constructing generalization bounds by introducing notions of local robustness and stability. The input space is partitioned into subregions, and the robustness term is scaled according to the proportion of stable versus unstable samples within each region, yielding a tighter, data- and model-dependent upper bound. This method is the first to refine global robustness measures into localized forms, substantially mitigating the vacuity commonly observed in traditional bounds. Experiments on ImageNet demonstrate that the proposed bound remains non-vacuous across various robust deep networks and closely tracks their empirical generalization errors, significantly outperforming existing approaches.

0 citationsRead paper

Advances in Battery Energy Storage Management: Control and Economic Synergies

Feb 06, 2026

This study addresses the common disconnect in existing battery energy storage system research between economic dispatch and operational control, which often compromises the dual objectives of grid stability and revenue maximization. By systematically reviewing five key areas—ancillary service modeling, real-time power control, optimization-based scheduling algorithms, techno-economic analysis, and digital twin technologies—the work proposes a synergistic framework that deeply integrates grid dispatch economics with battery control systems. The research identifies critical gaps in cross-domain coordination and establishes a digital twin architecture for energy storage that simultaneously supports high profitability and high reliability. This framework provides both theoretical foundations and practical technical pathways for the development of next-generation intelligent energy storage systems.

0 citationsRead paper

A Physics Informed Neural Network For Deriving MHD State Vectors From Global Active Regions Observations

Dec 23, 2025

Accurate early prediction of solar active regions (ARs) requires physically consistent initialization of the tachocline—the sub-photospheric magnetohydrodynamic (MHD) dynamo layer—yet conventional inversion methods lack physical self-consistency. Method: We propose PINNBARDS, a novel physics-informed neural network (PINN) framework that embeds the shallow-water MHD equations as hard constraints to enable end-to-end reconstruction of dynamically evolving, energy- and momentum-conserving tachocline initial states from observed photospheric magnetic ribbon geometry (e.g., SDO/HMI data). Contribution/Results: Applied to 14 February 2024 observations, PINNBARDS successfully reconstructed antisymmetric active band structures with optimal toroidal field strength of 20–30 kG and latitudinal width ≈10°, in excellent agreement with low-order radial-mode excitation theory. This constitutes the first physically grounded, observationally constrained initialization scheme enabling week-scale forecasting of AR emergence and associated flare activity.

0 citationsRead paper

A Validation Strategy for Deep Learning Models: Evaluating and Enhancing Robustness

Sep 23, 2025

Deep learning models exhibit insufficient robustness against adversarial perturbations and common image corruptions, undermining their reliability in real-world deployment. To address this, we propose an active robustness verification strategy that leverages the training set itself: by performing local robustness analysis, our method automatically identifies “weakly robust” samples—serving as early, interpretable indicators of model vulnerability—and enables targeted robustness enhancement. Unlike conventional passive paradigms that rely solely on perturbed test sets for robustness evaluation, ours is the first to repurpose training data for robustness diagnostics. We integrate adversarial perturbation injection with diverse natural corruption tests. Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet demonstrate that our strategy significantly improves model robustness against both attacks and corruptions (average gain of +8.2%) while enhancing the sensitivity and interpretability of reliability assessment.

0 citationsRead paper

Scalable Strategies for Continual Learning with Replay

May 18, 2025

Existing replay-based continual learning methods suffer from poor scalability and high computational overhead, while failing to effectively integrate parameter-efficient fine-tuning techniques—such as Low-Rank Adaptation (LoRA) and model merging—into multi-task adaptation. To address this, we propose a novel replay framework that deeply couples staged knowledge consolidation with sequential model merging tailored for task-ordered streams. It is the first to systematically integrate LoRA, task arithmetic, and collaborative optimization within a unified replay paradigm. Our approach reduces required replay sample volume by up to 55%, significantly improving training efficiency and model scalability. Empirically, it outperforms individual baseline methods across diverse multi-task sequences, achieving organic synergy between replay mechanisms and parameter-efficient adaptation. This work establishes a new paradigm for lightweight, accumulative continual learning models.

0 citationsRead paper