arXiv subjects

Kun Gao

Publications and source records attributed to Kun Gao.

At least 19 recordsLinked to original sources

arXiv2026

One Framework for All: Cross-Modal Membership Inference for Generative Models

Large generative models across text-to-text, text-to-image, and image-to-text modalities have been shown to pose significant privacy risks. One fundamental threat is membership inference attacks (MIA), which aim to determine whether a given data point was used in a model's training set. Although prior work has investigated MIAs against these three classes of generative models, existing approaches treat them in isolation and are not cross-applicable, thereby limiting their real-world utility. To address this limitation, we present the first comprehensive study of a unified membership inference framework that applies across text-to-text, text-to-image, and image-to-text modalities. Our approach is grounded in a key modality-agnostic observation: the output distribution of a generative model can approximate its training data distribution. Leveraging this property, we model the distributions of model-generated outputs and auxiliary non-member samples in a shared embedding space, and perform membership inference via likelihood ratio testing. We conduct extensive experiments in a strict black-box setting under both partial-knowledge and zero-knowledge threat models, and evaluate membership inference against both fine-tuning and pre-training data. Experimental results demonstrate our approach's superior performance in comparison to existing state-of-the-art methods, which are typically optimized for a single model class.

cs.LG↗

arXiv2026

UST-Hand: An Uncertainty-aware Spatiotemporal Point Cloud Interaction Network for 3D Self-supervised Hand Pose Estimation

Manually annotating accurate 3D hand poses is extremely time-consuming and labor-intensive. Existing self-supervised hand pose estimation methods leverage the discrepancy between input images and rendered outputs, or multi-view consistency constraints, as the driving force to optimize networks and progressively refine pose accuracy. However, these methods are highly susceptible to noisy pseudo-labels and overlook the importance of fully exploiting fine-grained spatial correlations, which undermines the stability of model training. To address these issues, we propose UST-Hand, a self-supervised learning framework that estimates uncertainty distribution of hand pose and constructs a probabilistic point cloud feature space, which enables the complex spatiotemporal relationship modeling. UST-Hand employs a conditional normalizing flow model to capture hand pose distributions and samples diverse hypotheses, facilitating robust learning under noisy pseudo-labels supervision with enhanced stability. These multi-hypothesis are mapped to a unified probabilistic 3D point cloud space for multi-view and temporal feature interaction, comprehensively exploring hand motion patterns and fine-grained spatial correlations. Extensive experiments on three challenging datasets demonstrate that UST-Hand achieves state-of-the-art performance, outperforming existing self-supervised methods by up to 37.8% in Mean Per Vertex Position Error (MPVPE).

cs.CV↗

arXiv2026

Visual Perceptual to Conceptual First-Order Rule Learning Networks

Learning rules plays a crucial role in deep learning, particularly in explainable artificial intelligence and enhancing the reasoning capabilities of large language models. While existing rule learning methods are primarily designed for symbolic data, learning rules from image data without supporting image labels and automatically inventing predicates remains a challenge. In this paper, we tackle these inductive rule learning problems from images with a framework called {\gamma}ILP, which provides a fully differentiable pipeline from image constant substitution to rule structure induction. Extensive experiments demonstrate that {\gamma}ILP achieves strong performance not only on classical symbolic relational datasets but also on relational image data and pure image datasets, such as Kandinsky patterns.

cs.AI↗

arXiv2026

Differentiable Rule Induction from Raw Sequence Inputs

Rule learning-based models are widely used in highly interpretable scenarios due to their transparent structures. Inductive logic programming (ILP), a form of machine learning, induces rules from facts while maintaining interpretability. Differentiable ILP models enhance this process by leveraging neural networks to improve robustness and scalability. However, most differentiable ILP methods rely on symbolic datasets, facing challenges when learning directly from raw data. Specifically, they struggle with explicit label leakage: The inability to map continuous inputs to symbolic variables without explicit supervision of input feature labels. In this work, we address this issue by integrating a self-supervised differentiable clustering model with a novel differentiable ILP model, enabling rule learning from raw data without explicit label leakage. The learned rules effectively describe raw data through its features. We demonstrate that our method intuitively and precisely learns generalized rules from time series and image data.

cs.AI↗

arXiv2025

R-Tuning: Wavelet-Decomposed Replay and Semantic Alignment for Continual Adaptation of Pretrained Time-Series Models

Pre-trained models have demonstrated exceptional generalization capabilities in time-series forecasting; however, adapting them to evolving data distributions remains a significant challenge. A key hurdle lies in accessing the original training data, as fine-tuning solely on new data often leads to catastrophic forgetting. To address this issue, we propose Replay Tuning (R-Tuning), a novel framework designed for the continual adaptation of pre-trained time-series models. R-Tuning constructs a unified latent space that captures both prior and current task knowledge through a frequency-aware replay strategy. Specifically, it augments model-generated samples via wavelet-based decomposition across multiple frequency bands, generating trend-preserving and fusion-enhanced variants to improve representation diversity and replay efficiency. To further reduce reliance on synthetic samples, R-Tuning introduces a latent consistency constraint that aligns new representations with the prior task space. This constraint guides joint optimization within a compact and semantically coherent latent space, ensuring robust knowledge retention and adaptation. Extensive experimental results demonstrate the superiority of R-Tuning, which reduces MAE and MSE by up to 46.9% and 46.8%, respectively, on new tasks, while preserving prior knowledge with gains of up to 5.7% and 6.0% on old tasks. Notably, under few-shot settings, R-Tuning outperforms all state-of-the-art baselines even when synthetic proxy samples account for only 5% of the new task dataset.

cs.LG↗

arXiv2025

Collaborative Learning-Enhanced Lightweight Models for Predicting Arterial Blood Pressure Waveform in a Large-scale Perioperative Dataset

Noninvasive arterial blood pressure (ABP) monitoring is essential for patient management in critical care and perioperative settings, providing continuous assessment of cardiovascular hemodynamics with minimal risks. Numerous deep learning models have developed to reconstruct ABP waveform from noninvasively acquired physiological signals such as electrocardiogram and photoplethysmogram. However, limited research has addressed the issue of model performance and computational load for deployment on embedded systems. The study introduces a lightweight sInvResUNet, along with a collaborative learning scheme named KDCL_sInvResUNet. With only 0.89 million parameters and a computational load of 0.02 GFLOPS, real-time ABP estimation was successfully achieved on embedded devices with an inference time of just 8.49 milliseconds for a 10-second output. We performed subject-independent validation in a large-scale and heterogeneous perioperative dataset containing 1,257,141 data segments from 2,154 patients, with a wide BP range (41-257 mmHg for SBP, and 31-234 mmHg for DBP). The proposed KDCL_sInvResUNet achieved lightly better performance compared to large models, with a mean absolute error of 10.06 mmHg and mean Pearson correlation of 0.88 in tracking ABP changes. Despite these promising results, all deep learning models showed significant performance variations across different demographic and cardiovascular conditions, highlighting their limited ability to generalize across such a broad and diverse population. This study lays a foundation work for real-time, unobtrusive ABP monitoring in real-world perioperative settings, providing baseline for future advancements in this area.

cs.LG↗

arXiv2025

Enhanced superconductivity in X4H15compounds via hole-doping at ambient pressure

This study presents a computational investigation of X4H15 compounds (where X represents a metal) as potential superconductors at ambient conditions or under pressure. Through systematic density functional theory calculations and electron-phonon coupling analysis, we demonstrate that electronic structure engineering via hole doping dramatically enhances the superconducting properties of these materials. While electron-doped compounds with X4+ cations (Ti, Zr, Hf, Th) exhibit modest transition temperatures of 1-9 K, hole-doped systems with X3+cations (Y, Tb, Dy, Ho,Er, Tm, Lu) show remarkably higher values of approximately 50 K at ambient pressure. Superconductivity in hole-doped compounds originates from stronger coupling between electrons and both cation and hydrogen phonon modes. Although pristine X3+4H15compounds are thermodynamically unstable, we propose a viable synthesis route via controlled hole doping of the charge-compensated YZr3H15 compound. Our calculations predict that even minimal concentrations of excess Y could induce high-temperature superconductivity while preserving structural integrity. This work reveals how strategic electronic structure modulation can optimize superconducting properties in hydride systems, establishing a promising pathway toward practical high-temperature conventional super-conductors at ambient pressure

cond-mat.supr-con↗

arXiv2025

The Maximum $T_c$ of Conventional Superconductors at Ambient Pressure

The theoretical maximum critical temperature ($T_c$) for conventional superconductors at ambient pressure remains a fundamental question in condensed matter physics. Through analysis of electron-phonon calculations for over 20,000 metals, we critically examine this question. We find that while hydride metals can exhibit maximum phonon frequencies of more than 5000 K, the crucial logarithmic average frequency $\omega_\text{log}$ rarely exceeds 1800 K. Our data reveals an inherent trade-off between $\omega_\text{log}$ and the electron-phonon coupling constant $\lambda$, suggesting that the optimal Eliashberg function that maximizes $T_c$ is unphysical. Based on our calculations, we identify Li$_2$AgH$_6$ and its sibling Li$_2$AuH$_6$ as theoretical materials that likely approach the practical limit for conventional superconductivity at ambient pressure. Analysis of thermodynamic stability indicates that compounds with higher predicted $T_c$ values are increasingly unstable, making their synthesis challenging. While fundamental physical laws do not strictly limit $T_c$ to low-temperatures, our analysis suggests that achieving room-temperature conventional superconductivity at ambient pressure is extremely unlikely.

cond-mat.supr-con↗

arXiv2025

Data Duplication: A Novel Multi-Purpose Attack Paradigm in Machine Unlearning

Duplication is a prevalent issue within datasets. Existing research has demonstrated that the presence of duplicated data in training datasets can significantly influence both model performance and data privacy. However, the impact of data duplication on the unlearning process remains largely unexplored. This paper addresses this gap by pioneering a comprehensive investigation into the role of data duplication, not only in standard machine unlearning but also in federated and reinforcement unlearning paradigms. Specifically, we propose an adversary who duplicates a subset of the target model's training set and incorporates it into the training set. After training, the adversary requests the model owner to unlearn this duplicated subset, and analyzes the impact on the unlearned model. For example, the adversary can challenge the model owner by revealing that, despite efforts to unlearn it, the influence of the duplicated subset remains in the model. Moreover, to circumvent detection by de-duplication techniques, we propose three novel near-duplication methods for the adversary, each tailored to a specific unlearning paradigm. We then examine their impacts on the unlearning process when de-duplication techniques are applied. Our findings reveal several crucial insights: 1) the gold standard unlearning method, retraining from scratch, fails to effectively conduct unlearning under certain conditions; 2) unlearning duplicated data can lead to significant model degradation in specific scenarios; and 3) meticulously crafted duplicates can evade detection by de-duplication methods.

cs.CR↗

arXiv2024

Prediction of high-Tc superconductivity in ternary actinium beryllium hydrides at low pressure

Hydrogen-rich superconductors are promising candidates to achieve room-temperature superconductivity. However, the extreme pressures needed to stabilize these structures significantly limit their practical applications. An effective strategy to reduce the external pressure is to add a light element M that binds with H to form MHx units, acting as a chemical precompressor. We exemplify this idea by performing ab initio calculations of the Ac-Be-H phase diagram, proving that the metallization pressure of Ac-H binaries, for which critical temperatures as high as 200 K were predicted at 200 GPa, can be significantly reduced via beryllium incorporation. We identify three thermodynamically stable (AcBe2H10, AcBeH8, and AcBe2H14) and four metastable compounds (fcc AcBeH8, AcBeH10, AcBeH12 and AcBe2H16). All of them are superconductors. In particular, fcc AcBeH8 remains dynamically stable down to 10 GPa, where it exhibits a superconducting transition temperature Tc of 181 K. The Be-H bonds are responsible for the exceptional properties of these ternary compounds and allow them to remain dynamically stable close to ambient pressure. Our results suggest that high-Tc superconductivity in hydrides is achievable at low pressure and may stimulate experimental synthesis of ternary hydrides.

cond-mat.supr-con↗

arXiv2023

Reinforcement Unlearning

Machine unlearning refers to the process of mitigating the influence of specific training data on machine learning models based on removal requests from data owners. However, one important area that has been largely overlooked in the research of unlearning is reinforcement learning. Reinforcement learning focuses on training an agent to make optimal decisions within an environment to maximize its cumulative rewards. During the training, the agent tends to memorize the features of the environment, which raises a significant concern about privacy. As per data protection regulations, the owner of the environment holds the right to revoke access to the agent's training data, thus necessitating the development of a novel and pressing research field, known as \emph{reinforcement unlearning}. Reinforcement unlearning focuses on revoking entire environments rather than individual data samples. This unique characteristic presents three distinct challenges: 1) how to propose unlearning schemes for environments; 2) how to avoid degrading the agent's performance in remaining environments; and 3) how to evaluate the effectiveness of unlearning. To tackle these challenges, we propose two reinforcement unlearning methods. The first method is based on decremental reinforcement learning, which aims to erase the agent's previously acquired knowledge gradually. The second method leverages environment poisoning attacks, which encourage the agent to learn new, albeit incorrect, knowledge to remove the unlearning environment. Particularly, to tackle the third challenge, we introduce the concept of ``environment inference attack'' to evaluate the unlearning outcomes.

cs.CR↗

arXiv2023

Probabilistic Prediction of Longitudinal Trajectory Considering Driving Heterogeneity with Interpretability

Automated vehicles are envisioned to navigate safely in complex mixed-traffic scenarios alongside human-driven vehicles. To promise a high degree of safety, accurately predicting the maneuvers of surrounding vehicles and their future positions is a critical task and attracts much attention. However, most existing studies focused on reasoning about positional information based on objective historical trajectories without fully considering the heterogeneity of driving behaviors. Therefore, this study proposes a trajectory prediction framework that combines Mixture Density Networks (MDN) and considers the driving heterogeneity to provide probabilistic and personalized predictions. Specifically, based on a certain length of historical trajectory data, the situation-specific driving preferences of each driver are identified, where key driving behavior feature vectors are extracted to characterize heterogeneity in driving behavior among different drivers. With the inputs of the short-term historical trajectory data and key driving behavior feature vectors, a probabilistic LSTMMD-DBV model combined with LSTM-based encoder-decoder networks and MDN layers is utilized to carry out personalized predictions. Finally, the SHapley Additive exPlanations (SHAP) method is employed to interpret the trained model for predictions. The proposed framework is tested based on a wide-range vehicle trajectory dataset. The results indicate that the proposed model can generate probabilistic future trajectories with remarkably improved predictions compared to existing benchmark models. Moreover, the results confirm that the additional input of driving behavior feature vectors representing the heterogeneity of driving behavior could provide more information and thus contribute to improving the prediction accuracy.

cs.LG↗

arXiv2023

Joint Sparse Representations and Coupled Dictionary Learning in Multi-Source Heterogeneous Image Pseudo-color Fusion

Considering that Coupled Dictionary Learning (CDL) method can obtain a reasonable linear mathematical relationship between resource images, we propose a novel CDL-based Synthetic Aperture Radar (SAR) and multispectral pseudo-color fusion method. Firstly, the traditional Brovey transform is employed as a pre-processing method on the paired SAR and multispectral images. Then, CDL is used to capture the correlation between the pre-processed image pairs based on the dictionaries generated from the source images via enforced joint sparse coding. Afterward, the joint sparse representation in the pair of dictionaries is utilized to construct an image mask via calculating the reconstruction errors, and therefore generate the final fusion image. The experimental verification results of the SAR images from the Sentinel-1 satellite and the multispectral images from the Landsat-8 satellite show that the proposed method can achieve superior visual effects, and excellent quantitative performance in terms of spectral distortion, correlation coefficient, MSE, NIQE, BRISQUE, and PIQE.

eess.IV↗

arXiv2023

Reinforcement Logic Rule Learning for Temporal Point Processes

We propose a framework that can incrementally expand the explanatory temporal logic rule set to explain the occurrence of temporal events. Leveraging the temporal point process modeling and learning framework, the rule content and weights will be gradually optimized until the likelihood of the observational event sequences is optimal. The proposed algorithm alternates between a master problem, where the current rule set weights are updated, and a subproblem, where a new rule is searched and included to best increase the likelihood. The formulated master problem is convex and relatively easy to solve using continuous optimization, whereas the subproblem requires searching the huge combinatorial rule predicate and relationship space. To tackle this challenge, we propose a neural search policy to learn to generate the new rule content as a sequence of actions. The policy parameters will be trained end-to-end using the reinforcement learning framework, where the reward signals can be efficiently queried by evaluating the subproblem objective. The trained policy can be used to generate new rules in a controllable way. We evaluate our methods on both synthetic and real healthcare datasets, obtaining promising results.

cs.LG↗

arXiv2023

Deciphering Spatial and Multi-scale Variations in the Effects of Key Factors of Maritime Safety: A Multi-scale Geographically Weighted Approach

Maritime accidents and corresponding consequences vary substantially across spatial dimensions as affected by various factors. Understanding the effects of key factors on maritime accident consequence would be of great benefit to prevent the occurrence or reduce the consequences of maritime accidents. Based on unique maritime accident data with geographical information covering fifteen years in the East China Sea, a multi-scale geographically weighted regression (MGWR) model considering the multi-scale spatial variation is employed to quantify the influences of different factors as well as the spatial heterogeneity in the effects of key factors on maritime accident consequence. The performances of MGWR are compared with multiple linear regression (MLR) and geographically weighted regression (GWR). Especially, MGWR outperforms the other two models in terms of modeling fitness and clearly capturing the unobserved spatial heterogeneity in effects of factors. Results reveal notably distinct and even inverse influences of some factors in different water areas on maritime accident consequences. For instance, approximately 50% of the accident locations present positive coefficients of good visibility while other locations are negative, which are ignored by MLR. The outcomes provide insights for making appropriate safety countermeasures and policies customized for different geographic areas.

physics.soc-ph↗

arXiv2022

Automatic detection of multilevel communities: scalable and resolution-limit-free

Community structure is one of the most important features of complex networks. Modularity-based methods for community detection typically rely on heuristic algorithms to optimize a specific community quality function. Such methods are limited by two major defects: (1) the resolution limit problem, which prohibits communities of heterogeneous sizes being simultaneously detected, and (2) divergent outputs of the heuristic algorithm, which make it difficult to differentiate relevant and irrelevant results. In this paper, we propose an improved method for community detection based on a scalable community "fitness function." We introduced a new parameter to enhance its scalability, and a strict strategy to filter the outputs. Due to the scalability, on the one hand our method is free of the resolution limit problem and performs excellently on large heterogeneous networks, while on the other hand it is capable of detecting more levels of communities than previous methods in deep hierarchical networks. Moreover, our strict strategy automatically removes redundant and irrelevant results, without any artificial selection. As a result, our method neatly outputs only the stable and unique communities, which are largely interpretable by the a priori knowledge about the network, including the implanted structures within synthetic networks, or metadata for real-world networks.

physics.soc-ph↗

arXiv2022

Learning First-Order Rules with Differentiable Logic Program Semantics

Learning first-order logic programs (LPs) from relational facts which yields intuitive insights into the data is a challenging topic in neuro-symbolic research. We introduce a novel differentiable inductive logic programming (ILP) model, called differentiable first-order rule learner (DFOL), which finds the correct LPs from relational facts by searching for the interpretable matrix representations of LPs. These interpretable matrices are deemed as trainable tensors in neural networks (NNs). The NNs are devised according to the differentiable semantics of LPs. Specifically, we first adopt a novel propositionalization method that transfers facts to NN-readable vector pairs representing interpretation pairs. We replace the immediate consequence operator with NN constraint functions consisting of algebraic operations and a sigmoid-like activation function. We map the symbolic forward-chained format of LPs into NN constraint functions consisting of operations between subsymbolic vector representations of atoms. By applying gradient descent, the trained well parameters of NNs can be decoded into precise symbolic LPs in forward-chained logic format. We demonstrate that DFOL can perform on several standard ILP datasets, knowledge bases, and probabilistic relation facts and outperform several well-known differentiable ILP models. Experimental results indicate that DFOL is a precise, robust, scalable, and computationally cheap differentiable ILP model.

cs.AI↗

arXiv2022

Divergent Effects of Factors on Crashes under Autonomous and Conventional Driving Modes Using A Hierarchical Bayesian Approach

Influencing factors on crashes involved with autonomous vehicles (AVs) have been paid increasing attention. However, there is a lack of comparative analyses between influencing factors on crashes of AVs and human-driven vehicles. To fill this research gap, the study aims to explore the divergent effects of factors on crashes under autonomous and conventional driving modes. This study obtained 154 publicly available autonomous vehicle crash data (70 for the autonomous driving mode and 84 for the conventional driving mode), and 36 explanatory variables were extracted from three categories, including environment, roads, and vehicles. Then, a hierarchical Bayesian approach was applied to analyze the impacting factors on crash type and severity under both driving modes. The results showed that some factors affected both driving modes, but their degrees were different. For example, the presence of turning movement had a greater impact on the crash severity under the conventional driving mode, while the presence of turning movement led to a larger decrease in the likelihood of rear-end crashes under the autonomous driving mode. More influencing factors only had a significant impact on one of the driving modes. For example, in the autonomous driving mode, two sidewalks decreased the severity of crashes, and on-street parking was positively associated with rear-end crashes, but they were not significant in the conventional driving mode. This study could contribute to the understanding and development of autonomous driving systems and the better coordination between autonomous driving and conventional driving.

stat.AP↗