arXiv subjects

Feng Chen

Publications and source records attributed to Feng Chen.

At least 163 records · Page 9Linked to original sources

arXiv2025

Evidential Uncertainty Probes for Graph Neural Networks

Accurate quantification of both aleatoric and epistemic uncertainties is essential when deploying Graph Neural Networks (GNNs) in high-stakes applications such as drug discovery and financial fraud detection, where reliable predictions are critical. Although Evidential Deep Learning (EDL) efficiently quantifies uncertainty using a Dirichlet distribution over predictive probabilities, existing EDL-based GNN (EGNN) models require modifications to the network architecture and retraining, failing to take advantage of pre-trained models. We propose a plug-and-play framework for uncertainty quantification in GNNs that works with pre-trained models without the need for retraining. Our Evidential Probing Network (EPN) uses a lightweight Multi-Layer-Perceptron (MLP) head to extract evidence from learned representations, allowing efficient integration with various GNN architectures. We further introduce evidence-based regularization techniques, referred to as EPN-reg, to enhance the estimation of epistemic uncertainty with theoretical justifications. Extensive experiments demonstrate that the proposed EPN-reg achieves state-of-the-art performance in accurate and efficient uncertainty quantification, making it suitable for real-world deployment.

cs.LG↗

arXiv2025

BIAN: A Deep Learning Method to Solve Inverse Problems Using Only Boundary Information

Over the past years, inverse problems in partial differential equations have garnered increasing interest among scientists and engineers. However, due to the lack of conventional stability, nonlinearity and non-convexity, these problems are quite challenging and difficult to solve. In this work, we propose a new kind of neural network to solve the coefficient identification problems with only the boundary information. In this work, three networks has been utilized as an approximator, a generator and a discriminator, respectively. This method is particularly useful in scenarios where the coefficients of interest have a complicated structure or are difficult to represent with traditional models. Comparative analysis against traditional coefficient estimation techniques demonstrates the superiority of our approach, not only handling highdimensional data and complex coefficient distributions adeptly by incorporating neural networks but also eliminating the necessity for extensive internal information due to the relationship between the energy distribution within the domain to the energy flux on the boundary. Several numerical examples have been presented to substantiate the merits of this algorithm including solving the Poisson equation and Helmholtz equation with spatially varying and piecewise uniform medium.

math.NA↗

arXiv2025

Global physics-informed neural networks (GPINNs): from local point-wise constraint to global nodal association

Recently, physics-informed neural networks (PINNs) and their variants have gained significant popularity as a scientific computing method for solving partial differential equations (PDEs), whereas accuracy is still its main shortcoming. Despite numerous development efforts, there is no literature demonstrating that these methods surpass classic numerical algorithms in solving the forward issue. In this paper, by analyzing the disparities between PINNs and traditional numerical methods based on mesh discretization, we investigate the underlying causes for the in adequate precision of PINNs and introduce a novel approach named global physics-informed neural networks (GPINNs). Inspired by the crucial concept of global nodal association in conventional numerical algorithms, GPINNs leverages the prior field distribution information from pre-trained PINNs to estimate the association weights between arbitrary nodes in space. GPINNs can not only be regarded as a meshless approach but also be demonstrated, both theoretically and in practical circumstances, to have the ability of second-order convergence when trained with equidistant nodes. Overall, GPINNs may be seen as an ideal approach to inheriting the merits of scientific machine learning (SciML) and conventional numerical computing, which also represent the first SciML algorithm to surpass standard numerical methods in terms of accuracy.

physics.comp-ph↗

arXiv2025

Predictive Uncertainty Quantification for Bird's Eye View Segmentation: A Benchmark and Novel Loss Function

The fusion of raw sensor data to create a Bird's Eye View (BEV) representation is critical for autonomous vehicle planning and control. Despite the growing interest in using deep learning models for BEV semantic segmentation, anticipating segmentation errors and enhancing the explainability of these models remain underexplored. This paper introduces a comprehensive benchmark for predictive uncertainty quantification in BEV segmentation, evaluating multiple uncertainty quantification methods across three popular datasets with three representative network architectures. Our study focuses on the effectiveness of quantified uncertainty in detecting misclassified and out-of-distribution (OOD) pixels while also improving model calibration. Through empirical analysis, we uncover challenges in existing uncertainty quantification methods and demonstrate the potential of evidential deep learning techniques, which capture both aleatoric and epistemic uncertainty. To address these challenges, we propose a novel loss function, Uncertainty-Focal-Cross-Entropy (UFCE), specifically designed for highly imbalanced data, along with a simple uncertainty-scaling regularization term that improves both uncertainty quantification and model calibration for BEV segmentation.

cs.LG↗

arXiv2025

Global Phase Diagram of D-wave Superconductivity in the Square-Lattice $t\text{-}J$ Model

The Hubbard and closely related $t\text-J$ models are exciting platforms for unconventional superconductivity (SC). Through state-of-the-art density matrix renormalization group calculations using the grand canonical ensemble, we address open issues regarding the ground-state phase diagram of the extended $t\text{-}J$ model on a square lattice in the parameter regime relevant to cuprate superconductors. On large 8-leg cylinders, we demonstrate that the pure $t\text{-}J$ model with only nearest-neighbor hoppings and superexchange interactions, for a wide range of doping ($δ=0.1-0.2$), hosts robust d-wave superconductivity possibly coexisting with weak unidirectional pair density wave. Furthermore, a small next nearest neighbor hopping $t_2$ suppresses pair and charge density waves, resulting in a uniform d-wave SC phase in both electron- and hole-doped cuprate model systems. Our work validates the $t\text-J$ model as a proper minimum model for the emergence of superconductivity in cuprate superconductors.

cond-mat.supr-con↗

arXiv2025

Robust non-Abelian even-denominator fractional Chern insulator in twisted bilayer MoTe$_2$

A recent experiment observes a series of quantum spin Hall effects in transition metal dichalcogenide moiré MoTe$_2$ [K. Kang, et. al, Nature 628, 522-526 (2024)]. Among them, the vanishing Hall signal at the filling factor $ν=3$ implies a possible realization of a time-reversal pair of the even-denominator fractional Chern insulators (FCIs). Inspired by this discovery, we investigate whether a robust incompressible quantum Hall liquid can be stabilized in the half-filled Chern band of twisted MoTe$_2$ bilayers. We use the continuum model with parameters relevant to twisted MoTe$_2$ bilayers and obtain three consecutive nearly flat Chern bands with the same Chern number. Crucially, when the second moiré miniband is half-filled,signatures of non-Abelian frctional quantum Hall state are found via exact diagonalization calculations, including the stable six-fold ground state degeneracy that grows more robust with the lattice size and is consistent with an even-denominator FCI state. We further perform flux insertion simulations to reveal a 1/2 quantized many-body Chern number as direct evidence of topological order. Furthermore, the ground state density structure factors show no sharp peak, indicating no charge density wave order. These evidences signal the potential of realizing the non-Abelian state at zero magnetic field in twisted bilayer MoTe$_2$ at the fractional hole filling 3/2.

cond-mat.str-el↗

arXiv2025

Supersonic flow kinetics: Mesoscale structures, thermodynamic nonequilibrium effects and entropy production mechanisms

Supersonic flow is a typical nonlinear, nonequilibrium, multiscale, and complex phenomenon. This paper applies discrete Boltzmann method/model (DBM) to simulate and analyze these characteristics. A Burnett-level DBM for supersonic flow is constructed based on the Shakhov-BGK model. Higher-order analytical expressions for thermodynamic nonequilibrium effects are derived, providing a constitutive basis for improving traditional macroscopic hydrodynamics modeling. Criteria for evaluating the validity of DBM are established by comparing numerical and analytical solutions of nonequilibrium measures. The multiscale DBM is used to investigate discrete/nonequilibrium characteristics and entropy production mechanisms in shock regular reflection. The findings include: (a) Compared to NS-level DBM, the Burnett-level DBM offers more accurate representations of viscous stress and heat flux, ensures non-negativity of entropy production in accordance with the second law of thermodynamics, and exhibits better numerical stability. (b) Near the interfaces of incident and reflected shock waves, strong nonequilibrium driving forces lead to prominent nonequilibrium effects. By monitoring the timing and location of peak nonequilibrium quantities, the evolution characteristics of incident and reflected shock waves can be accurately and dynamically tracked. (c) In the intermediate state, the bent reflected shock and incident shock interface are wider and exhibit lower nonequilibrium intensities compared to their final state. (d) The Mach number enhances various kinds of nonequilibrium intensities in a power-law manner $D_{mn} \sim \mathtt{Ma}^α$. The power exponent $α$ and kinetic modes of nonequilibrium effects $m$ follows a logarithmic relation $α\sim \ln (m - m_0)$. This research provides new perspectives and kinetic insights into supersonic flow studies.

physics.flu-dyn↗

arXiv2025

A method of supervised learning from conflicting data with hidden contexts

Conventional supervised learning assumes a stable input-output relationship. However, this assumption fails in open-ended training settings where the input-output relationship depends on hidden contexts. In this work, we formulate a more general supervised learning problem in which training data is drawn from multiple unobservable domains, each potentially exhibiting distinct input-output maps. This inherent conflict in data renders standard empirical risk minimization training ineffective. To address this challenge, we propose a method LEAF that introduces an allocation function, which learns to assign conflicting data to different predictive models. We establish a connection between LEAF and a variant of the Expectation-Maximization algorithm, allowing us to derive an analytical expression for the allocation function. Finally, we provide a theoretical analysis of LEAF and empirically validate its effectiveness on both synthetic and real-world tasks involving conflicting data.

cs.LG↗

arXiv2025

Feature contamination: Neural networks learn uncorrelated features and fail to generalize

Learning representations that generalize under distribution shifts is critical for building robust machine learning models. However, despite significant efforts in recent years, algorithmic advances in this direction have been limited. In this work, we seek to understand the fundamental difficulty of out-of-distribution generalization with deep neural networks. We first empirically show that perhaps surprisingly, even allowing a neural network to explicitly fit the representations obtained from a teacher network that can generalize out-of-distribution is insufficient for the generalization of the student network. Then, by a theoretical study of two-layer ReLU networks optimized by stochastic gradient descent (SGD) under a structured feature model, we identify a fundamental yet unexplored feature learning proclivity of neural networks, feature contamination: neural networks can learn uncorrelated features together with predictive features, resulting in generalization failure under distribution shifts. Notably, this mechanism essentially differs from the prevailing narrative in the literature that attributes the generalization failure to spurious correlations. Overall, our results offer new insights into the non-linear feature learning dynamics of neural networks and highlight the necessity of considering inductive biases in out-of-distribution generalization.

cs.LG↗

arXiv2025

GMT: Guided Mask Transformer for Leaf Instance Segmentation

Leaf instance segmentation is a challenging multi-instance segmentation task, aiming to separate and delineate each leaf in an image of a plant. Accurate segmentation of each leaf is crucial for plant-related applications such as the fine-grained monitoring of plant growth and crop yield estimation. This task is challenging because of the high similarity (in shape and colour), great size variation, and heavy occlusions among leaf instances. Furthermore, the typically small size of annotated leaf datasets makes it more difficult to learn the distinctive features needed for precise segmentation. We hypothesise that the key to overcoming the these challenges lies in the specific spatial patterns of leaf distribution. In this paper, we propose the Guided Mask Transformer (GMT), which leverages and integrates leaf spatial distribution priors into a Transformer-based segmentor. These spatial priors are embedded in a set of guide functions that map leaves at different positions into a more separable embedding space. Our GMT consistently outperforms the state-of-the-art on three public plant datasets. Our code is available at https://github.com/vios-s/gmt-leaf-ins-seg.

cs.CV↗

arXiv2025

Cross-Dataset Gaze Estimation by Evidential Inter-intra Fusion

Achieving accurate and reliable gaze predictions in complex and diverse environments remains challenging. Fortunately, it is straightforward to access diverse gaze datasets in real-world applications. We discover that training these datasets jointly can significantly improve the generalization of gaze estimation, which is overlooked in previous works. However, due to the inherent distribution shift across different datasets, simply mixing multiple dataset decreases the performance in the original domain despite gaining better generalization abilities. To address the problem of ``cross-dataset gaze estimation'', we propose a novel Evidential Inter-intra Fusion EIF framework, for training a cross-dataset model that performs well across all source and unseen domains. Specifically, we build independent single-dataset branches for various datasets where the data space is partitioned into overlapping subspaces within each dataset for local regression, and further create a cross-dataset branch to integrate the generalizable features from single-dataset branches. Furthermore, evidential regressors based on the Normal and Inverse-Gamma (NIG) distribution are designed to additionally provide uncertainty estimation apart from predicting gaze. Building upon this foundation, our proposed framework achieves both intra-evidential fusion among multiple local regressors within each dataset and inter-evidential fusion among multiple branches by Mixture \textbfof Normal Inverse-Gamma (MoNIG distribution. Experiments demonstrate that our method consistently achieves notable improvements in both source domains and unseen domains.

cs.CV↗

arXiv2024

ZipVL: Efficient Large Vision-Language Models with Dynamic Token Sparsification

The efficiency of large vision-language models (LVLMs) is constrained by the computational bottleneck of the attention mechanism during the prefill phase and the memory bottleneck of fetching the key-value (KV) cache in the decoding phase, particularly in scenarios involving high-resolution images or videos. Visual content often exhibits substantial redundancy, resulting in highly sparse attention maps within LVLMs. This sparsity can be leveraged to accelerate attention computation or compress the KV cache through various approaches. However, most studies focus on addressing only one of these bottlenecks and do not adequately support dynamic adjustment of sparsity concerning distinct layers or tasks. In this paper, we present ZipVL, an efficient inference framework designed for LVLMs through a dynamic ratio allocation strategy of important tokens. This ratio is adaptively determined based on the layer-specific distribution of attention scores, rather than fixed hyper-parameters, thereby improving efficiency for less complex tasks while maintaining high performance for more challenging ones. Then we select important tokens based on their normalized attention scores and perform sparse attention mechanism solely on those important tokens, reducing the latency in the prefill phase. Tokens deemed less important will be discarded to reduce KV cache size, alleviating the memory bottleneck in the decoding phase. Our experiments demonstrate that ZipVL can accelerate the prefill phase by 2.3$\times$ and improve decoding throughput by 2.8$\times$, with a minimal accuracy reduction of only 0.5\% on VQAv2 benchmark over LLaVA-Next-13B model, effectively enhancing the generation efficiency of LVLMs.

cs.CV↗

arXiv2024

High-efficiency On-chip Quantum Photon Source in Modal Phase-matched Lithium Niobate Nanowaveguide

Thin-film lithium niobate on insulator~(LNOI) emerges as a promising platform for integrated quantum photon source, enabling scalable on-chip quantum information processing. The most popular technique to overcome the phase mismatching between interacting waves in waveguide is periodic poling, which is intrinsically sensitive to poling uniformity. Here, we report an alternative strategy to offset the phase mismatching of spontaneous parametric down-conversion~(SPDC) process, so-called modal phase matching, in a straight waveguide fabricated on a dual-layer LNOI. The dual-layer LNOI consists of two 300~nm lithium niobates with opposite directions, which significantly enhances the spatial overlap between fundamental and high-order modes and thus enables efficient SPDC. This dual-layer waveguide generates photon pairs with pair generation rate of 41.77~GHz/mW, which exhibits excellent signal-to-noise performance with coincidence-to-accidental ratio up to 58298$\pm$1297. Moreover, we observe a heralded single-photon source with second-order autocorrelation $g_{H}^{(2)}(0)<0.2$ and heralded rate exceeding 100~kHz. Our results provide an experiment-friendly approach for efficient generation of quantum photon sources and benefit the on-chip quantum information processing based on LNOI.

quant-ph↗

arXiv2024

ZipCache: A DRAM/SSD Cache with Built-in Transparent Compression

As a core component in modern data centers, key-value cache provides high-throughput and low-latency services for high-speed data processing. The effectiveness of a key-value cache relies on its ability of accommodating the needed data. However, expanding the cache capacity is often more difficult than commonly expected because of many practical constraints, such as server costs, cooling issues, rack space, and even human resource expenses. A potential solution is compression, which virtually extends the cache capacity by condensing data in cache. In practice, this seemingly simple idea has not gained much traction in key-value cache system design, due to several critical issues: the compression-unfriendly index structure, severe read/write amplification, wasteful decompression operations, and heavy computing cost. This paper presents a hybrid DRAM-SSD cache design to realize a systematic integration of data compression in key-value cache. By treating compression as an essential component, we have redesigned the indexing structure, data management, and leveraged the emerging computational SSD hardware for collaborative optimizations. We have developed a prototype, called ZipCache. Our experimental results show that ZipCache can achieve up to 72.4% higher throughput and 42.4% lower latency, while reducing the write amplification by up to 26.2 times.

cs.DB↗

arXiv2024

Stable Continual Reinforcement Learning via Diffusion-based Trajectory Replay

Given the inherent non-stationarity prevalent in real-world applications, continual Reinforcement Learning (RL) aims to equip the agent with the capability to address a series of sequentially presented decision-making tasks. Within this problem setting, a pivotal challenge revolves around \textit{catastrophic forgetting} issue, wherein the agent is prone to effortlessly erode the decisional knowledge associated with past encountered tasks when learning the new one. In recent progresses, the \textit{generative replay} methods have showcased substantial potential by employing generative models to replay data distribution of past tasks. Compared to storing the data from past tasks directly, this category of methods circumvents the growing storage overhead and possible data privacy concerns. However, constrained by the expressive capacity of generative models, existing \textit{generative replay} methods face challenges in faithfully reconstructing the data distribution of past tasks, particularly in scenarios with a myriad of tasks or high-dimensional data. Inspired by the success of diffusion models in various generative tasks, this paper introduces a novel continual RL algorithm DISTR (Diffusion-based Trajectory Replay) that employs a diffusion model to memorize the high-return trajectory distribution of each encountered task and wakeups these distributions during the policy learning on new tasks. Besides, considering the impracticality of replaying all past data each time, a prioritization mechanism is proposed to prioritize the trajectory replay of pivotal tasks in our method. Empirical experiments on the popular continual RL benchmark \texttt{Continual World} demonstrate that our proposed method obtains a favorable balance between \textit{stability} and \textit{plasticity}, surpassing various existing continual RL baselines in average success rate.

cs.LG↗

arXiv2024

Feature-Space Semantic Invariance: Enhanced OOD Detection for Open-Set Domain Generalization

Open-set domain generalization addresses a real-world challenge: training a model to generalize across unseen domains (domain generalization) while also detecting samples from unknown classes not encountered during training (open-set recognition). However, most existing approaches tackle these issues separately, limiting their practical applicability. To overcome this limitation, we propose a unified framework for open-set domain generalization by introducing Feature-space Semantic Invariance (FSI). FSI maintains semantic consistency across different domains within the feature space, enabling more accurate detection of OOD instances in unseen domains. Additionally, we adopt a generative model to produce synthetic data with novel domain styles or class labels, enhancing model robustness. Initial experiments show that our method improves AUROC by 9.1% to 18.9% on ColoredMNIST, while also significantly increasing in-distribution classification accuracy.

cs.CV↗

arXiv2024

FEED: Fairness-Enhanced Meta-Learning for Domain Generalization

Generalizing to out-of-distribution data while being aware of model fairness is a significant and challenging problem in meta-learning. The goal of this problem is to find a set of fairness-aware invariant parameters of classifier that is trained using data drawn from a family of related training domains with distribution shift on non-sensitive features as well as different levels of dependence between model predictions and sensitive features so that the classifier can achieve good generalization performance on unknown but distinct test domains. To tackle this challenge, existing state-of-the-art methods either address the domain generalization problem but completely ignore learning with fairness or solely specify shifted domains with various fairness levels. This paper introduces an approach to fairness-aware meta-learning that significantly enhances domain generalization capabilities. Our framework, Fairness-Enhanced Meta-Learning for Domain Generalization (FEED), disentangles latent data representations into content, style, and sensitive vectors. This disentanglement facilitates the robust generalization of machine learning models across diverse domains while adhering to fairness constraints. Unlike traditional methods that focus primarily on domain invariance or sensitivity to shifts, our model integrates a fairness-aware invariance criterion directly into the meta-learning process. This integration ensures that the learned parameters uphold fairness consistently, even when domain characteristics vary widely. We validate our approach through extensive experiments across multiple benchmarks, demonstrating not only superior performance in maintaining high accuracy and fairness but also significant improvements over existing state-of-the-art methods in domain generalization tasks.

cs.LG↗

arXiv2024

MADOD: Generalizing OOD Detection to Unseen Domains via G-Invariance Meta-Learning

Real-world machine learning applications often face simultaneous covariate and semantic shifts, challenging traditional domain generalization and out-of-distribution (OOD) detection methods. We introduce Meta-learned Across Domain Out-of-distribution Detection (MADOD), a novel framework designed to address both shifts concurrently. MADOD leverages meta-learning and G-invariance to enhance model generalizability and OOD detection in unseen domains. Our key innovation lies in task construction: we randomly designate in-distribution classes as pseudo-OODs within each meta-learning task, simulating OOD scenarios using existing data. This approach, combined with energy-based regularization, enables the learning of robust, domain-invariant features while calibrating decision boundaries for effective OOD detection. Operating in a test domain-agnostic setting, MADOD eliminates the need for adaptation during inference, making it suitable for scenarios where test data is unavailable. Extensive experiments on real-world and synthetic datasets demonstrate MADOD's superior performance in semantic OOD detection across unseen domains, achieving an AUPR improvement of 8.48% to 20.81%, while maintaining competitive in-distribution classification accuracy, representing a significant advancement in handling both covariate and semantic shifts.

cs.LG↗