arXiv subjects

Feng Ji

Publications and source records attributed to Feng Ji.

At least 37 records · Page 2Linked to original sources

arXiv2025

Modulo Video Recovery via Selective Spatiotemporal Vision Transformer

Conventional image sensors have limited dynamic range, causing saturation in high-dynamic-range (HDR) scenes. Modulo cameras address this by folding incident irradiance into a bounded range, yet require specialized unwrapping algorithms to reconstruct the underlying signal. Unlike HDR recovery, which extends dynamic range from conventional sampling, modulo recovery restores actual values from folded samples. Despite being introduced over a decade ago, progress in modulo image recovery has been slow, especially in the use of modern deep learning techniques. In this work, we demonstrate that standard HDR methods are unsuitable for modulo recovery. Transformers, however, can capture global dependencies and spatial-temporal relationships crucial for resolving folded video frames. Still, adapting existing Transformer architectures for modulo recovery demands novel techniques. To this end, we present Selective Spatiotemporal Vision Transformer (SSViT), the first deep learning framework for modulo video reconstruction. SSViT employs a token selection strategy to improve efficiency and concentrate on the most critical regions. Experiments confirm that SSViT produces high-quality reconstructions from 8-bit folded videos and achieves state-of-the-art performance in modulo video recovery.

cs.CV↗

arXiv2025

Generative Adversarial Networks for High-Dimensional Item Factor Analysis: A Deep Adversarial Learning Algorithm

Advances in deep learning and representation learning have transformed item factor analysis (IFA) in the item response theory (IRT) literature by enabling more efficient and accurate parameter estimation. Variational Autoencoders (VAEs) have been one of the most impactful techniques in modeling high-dimensional latent variables in this context. However, the limited expressiveness of the inference model based on traditional VAEs can still hinder the estimation performance. We introduce Adversarial Variational Bayes (AVB) algorithms as an improvement to VAEs for IFA with improved flexibility and accuracy. By bridging the strengths of VAEs and Generative Adversarial Networks (GANs), AVB incorporates an auxiliary discriminator network to reframe the estimation process as a two-player adversarial game and removes the restrictive assumption of standard normal distributions in the inference model. Theoretically, AVB can achieve similar or higher likelihood compared to VAEs. A further enhanced algorithm, Importance-weighted Adversarial Variational Bayes (IWAVB) is proposed and compared with Importance-weighted Autoencoders (IWAE). In an exploratory analysis of empirical data, IWAVB demonstrated superior expressiveness by achieving a higher likelihood compared to IWAE. In confirmatory analysis with simulated data, IWAVB achieved similar mean-square error results to IWAE while consistently achieving higher likelihoods. When latent variables followed a multimodal distribution, IWAVB outperformed IWAE. With its innovative use of GANs, IWAVB is shown to have the potential to extend IFA to handle large-scale data, facilitating the potential integration of psychometrics and multimodal data analysis.

stat.ML↗

arXiv2025

Sampling Theory of Jointly Bandlimited Time-vertex Graph Signals

Time-vertex graph signal (TVGS) models describe time-varying data with irregular structures. The bandlimitedness in the joint time-vertex Fourier spectral domain reflects smoothness in both temporal and graph topology. In this paper, we study the critical sampling of three types of TVGS including continuous-time signals, infinite-length sequences, and finite-length sequences in the time domain for each vertex on the graph. For a jointly bandlimited TVGS, we prove a lower bound on sampling density or sampling ratio, which depends on the measure of the spectral support in the joint time-vertex Fourier spectral domain. We also provide a lower bound on the sampling density or sampling ratio of each vertex on sampling sets for perfect recovery. To demonstrate that critical sampling is achievable, we propose the sampling and reconstruction procedures for the different types of TVGS. Finally, we show how the proposed sampling schemes can be applied to numerical as well as real datasets.

eess.SP↗

arXiv2025

Omne-R1: Learning to Reason with Memory for Multi-hop Question Answering

This paper introduces Omne-R1, a novel approach designed to enhance multi-hop question answering capabilities on schema-free knowledge graphs by integrating advanced reasoning models. Our method employs a multi-stage training workflow, including two reinforcement learning phases and one supervised fine-tuning phase. We address the challenge of limited suitable knowledge graphs and QA data by constructing domain-independent knowledge graphs and auto-generating QA pairs. Experimental results show significant improvements in answering multi-hop questions, with notable performance gains on more complex 3+ hop questions. Our proposed training framework demonstrates strong generalization abilities across diverse knowledge domains.

cs.CL↗

arXiv2025

MiroMind-M1: An Open-Source Advancement in Mathematical Reasoning via Context-Aware Multi-Stage Policy Optimization

Large language models have recently evolved from fluent text generation to advanced reasoning across diverse domains, giving rise to reasoning language models. Among these domains, mathematical reasoning serves as a representative benchmark as it requires precise multi-step logic and abstract reasoning, which can be generalized to other tasks. While closed-source RLMs such as GPT-o3 demonstrate impressive reasoning capabilities, their proprietary nature limits transparency and reproducibility. Although many open-source projects aim to close this gap, most of them lack sufficient openness by omitting critical resources such as datasets and detailed training configurations, which hinders reproducibility. To contribute toward greater transparency in RLM development, we introduce the MiroMind-M1 series, a set of fully open-source RLMs built on the Qwen-2.5 backbone that match or exceed the performance of existing open-source RLMs. Specifically, our models are trained in two stages: SFT on a carefully curated corpus of 719K math-reasoning problems with verified CoT trajectories, followed by RLVR on 62K challenging and verifiable problems. To enhance the robustness and efficiency of the RLVR process, we introduce Context-Aware Multi-Stage Policy Optimization, an algorithm that integrates length-progressive training with an adaptive repetition penalty to encourage context-aware RL training. Our model achieves state-of-the-art or competitive performance and superior token efficiency among Qwen-2.5-based open-source 7B and 32B models on the AIME24, AIME25, and MATH benchmarks. To facilitate reproducibility, we release the complete stack: models (MiroMind-M1-SFT-7B, MiroMind-M1-RL-7B, MiroMind-M1-RL-32B); datasets (MiroMind-M1-SFT-719K, MiroMind-M1-RL-62K); and all training and evaluation configurations. We hope these resources will support further research and foster community advancement.

cs.CL↗

arXiv2025

A Generalized Graph Signal Processing Framework for Multiple Hypothesis Testing over Networks

We consider the multiple hypothesis testing (MHT) problem over the joint domain formed by a graph and a measure space. On each sample point of this joint domain, we assign a hypothesis test and a corresponding $p$-value. The goal is to make decisions for all hypotheses simultaneously, using all available $p$-values. In practice, this problem resembles the detection problem over a sensor network during a period of time. To solve this problem, we extend the traditional two-groups model such that the prior probability of the null hypothesis and the alternative distribution of $p$-values can be inhomogeneous over the joint domain. We model the inhomogeneity via a generalized graph signal. This more flexible statistical model yields a more powerful detection strategy by leveraging the information from the joint domain.

eess.SP↗

arXiv2025

Simple Graph Contrastive Learning via Fractional-order Neural Diffusion Networks

Graph Contrastive Learning (GCL) has recently made progress as an unsupervised graph representation learning paradigm. GCL approaches can be categorized into augmentation-based and augmentation-free methods. The former relies on complex data augmentations, while the latter depends on encoders that can generate distinct views of the same input. Both approaches may require negative samples for training. In this paper, we introduce a novel augmentation-free GCL framework based on graph neural diffusion models. Specifically, we utilize learnable encoders governed by Fractional Differential Equations (FDE). Each FDE is characterized by an order parameter of the differential operator. We demonstrate that varying these parameters allows us to produce learnable encoders that generate diverse views, capturing either local or global information, for contrastive learning. Our model does not require negative samples for training and is applicable to both homophilic and heterophilic datasets. We demonstrate its effectiveness across various datasets, achieving state-of-the-art performance.

cs.LG↗

arXiv2025

Helically symmetric solution of 3D Euler equations with vorticity and its free boundary

This paper investigates an incompressible steady free boundary problem of Euler equations with helical symmetry in $3$ dimensions and with nontrivial vorticity. The velocity field of the fluid arises from the spiral of its velocity within a cross-section, whose global existence, uniqueness and well-posedness with fixed boundary were established by a series of brilliant works. A perplexing issue, untouched in the literature, concerns the free boundary problem with (partial) unknown domain boundary in this helically symmetric configuration. We address this gap through the analysis of the optimal regularity property of the scalar stream function as a minimizer in a semilinear minimal problem, establishing the $C^{0,1}$-regularity of the minimizer, and the $C^{1,α}$-regularity of its free boundary. More specifically, the regularity results are obtained in arbitrary cross-sections through smooth helical transformation by virtue of variational method and the rule of "flatness implies $C^{1,α}$".

math.AP↗

arXiv2025

Falcon: Faster and Parallel Inference of Large Language Models through Enhanced Semi-Autoregressive Drafting and Custom-Designed Decoding Tree

Striking an optimal balance between minimal drafting latency and high speculation accuracy to enhance the inference speed of Large Language Models remains a significant challenge in speculative decoding. In this paper, we introduce Falcon, an innovative semi-autoregressive speculative decoding framework fashioned to augment both the drafter's parallelism and output quality. Falcon incorporates the Coupled Sequential Glancing Distillation technique, which fortifies inter-token dependencies within the same block, leading to increased speculation accuracy. We offer a comprehensive theoretical analysis to illuminate the underlying mechanisms. Additionally, we introduce a Custom-Designed Decoding Tree, which permits the drafter to generate multiple tokens in a single forward pass and accommodates multiple forward passes as needed, thereby boosting the number of drafted tokens and significantly improving the overall acceptance rate. Comprehensive evaluations on benchmark datasets such as MT-Bench, HumanEval, and GSM8K demonstrate Falcon's superior acceleration capabilities. The framework achieves a lossless speedup ratio ranging from 2.91x to 3.51x when tested on the Vicuna and LLaMA2-Chat model series. These results outstrip existing speculative decoding methods for LLMs, including Eagle, Medusa, Lookahead, SPS, and PLD, while maintaining a compact drafter architecture equivalent to merely two Transformer layers.

cs.CL↗

arXiv2025

Control the GNN: Utilizing Neural Controller with Lyapunov Stability for Test-Time Feature Reconstruction

The performance of graph neural networks (GNNs) is susceptible to discrepancies between training and testing sample distributions. Prior studies have attempted to mitigating the impact of distribution shift by reconstructing node features during the testing phase without modifying the model parameters. However, these approaches lack theoretical analysis of the proximity between predictions and ground truth at test time. In this paper, we propose a novel node feature reconstruction method grounded in Lyapunov stability theory. Specifically, we model the GNN as a control system during the testing phase, considering node features as control variables. A neural controller that adheres to the Lyapunov stability criterion is then employed to reconstruct these node features, ensuring that the predictions progressively approach the ground truth at test time. We validate the effectiveness of our approach through extensive experiments across multiple datasets, demonstrating significant performance improvements.

cs.LG↗

arXiv2025

Effective Numerical Simulation of Fault Transient System

Power systems, including synchronous generator systems, are typical systems that strive for stable operation. In this article, we numerically study the fault transient process of a synchronous generator system based on the first benchmark model. That is, we make it clear whether an originally stable generator system can restore its stability after a short time of unstable transient process. To achieve this, we construct a structure-preserving method and compare it with the existing and frequently-used predictor-corrector method. We newly establish a reductive form of the circuit system and accelerate the reduction process. Also a switching method between two stages in the fault transient process is given. Numerical results show the effectiveness and reliability of our method.

math.NA↗

arXiv2025

Singular and regular analysis for the free boundaries of two-phase inviscid fluids in gravity field

In this paper, we consider a free boundary problem of two-phase inviscid incompressible fluid in gravity field. The presence of the gravity field induces novel phenomena that there might be some stagnation points on free surface of the two-phase flow, where the velocity field of the fluid vanishes. From the mathematical point of view, the gradient of the stream function degenerates near the stagnation point, leading to singular behaviors on the free surface. The primary objective of this study is to investigate the singularity and regularity of the two-phase free surface, considering their mutual interaction between the two incompressible fluids in two dimensions. More precisely, if the two fluids meet locally at a single point, referred to as the possible two-phase stagnation point, we demonstrate that the singular side of the two-phase free surface exhibits a symmetric Stokes singular profile, while the regular side near this point maintains the $C^{1,α}$ regularity. On the other hand, if the free surfaces of the two fluids stick together and have non-trivial overlapping common boundary at the stagnation point, then the interaction between the two fluids will break the symmetry of the Stokes corner profile, which is attached to the $C^{1,α}$ regular free surface on the other side. As a byproduct of our analysis, it's shown that the velocity field for the two fluids cannot vanish simultaneously on the two-phase free boundary. Our results generalize the significant works on the Stokes conjecture in [V$\check{a}$rv$\check{a}$ruc$\check{a}$-Weiss, Acta Math., 206, (2011)] for one-phase gravity water wave, and on regular results on the free boundaries in [De Philippis-Spolaor-Velichkov, Invent. Math., 225, (2021)] for two-phase fluids without gravity.

math.AP↗

arXiv2025

A Graph Signal Processing Perspective of Network Multiple Hypothesis Testing with False Discovery Rate Control

We consider a multiple hypothesis testing problem in a sensor network over the joint spatio-temporal domain. The sensor network is modeled as a graph, with each vertex representing a sensor and a signal over time associated with each vertex. We assume a hypothesis test and an associated $p$-value for every sample point in the joint spatio-temporal domain. Our goal is to determine which points have true alternative. By parameterizing the unknown $p$-value distribution under the alternative and the prior probabilities of hypotheses being null with a bandlimited generalized graph signal, we can obtain consistent estimates for them. Consequently, we also obtain an estimate of the local false discovery rates (lfdr). We prove that by using a step-up procedure on the estimated lfdr, we can achieve asymptotic false discovery rate control at a pre-determined level. Numerical experiments validate the effectiveness of our approach compared to existing methods.

eess.SP↗

arXiv2024

Valid standard errors for Bayesian quantile regression with clustered and independent data

In Bayesian quantile regression, the most commonly used likelihood is the asymmetric Laplace (AL) likelihood. The reason for this choice is not that it is a plausible data-generating model but that the corresponding maximum likelihood estimator is identical to the classical estimator by Koenker and Bassett (1978), and in that sense, the AL likelihood can be thought of as a working likelihood. AL-based quantile regression has been shown to produce good finite-sample Bayesian point estimates and to be consistent. However, if the AL distribution does not correspond to the data-generating distribution, credible intervals based on posterior standard deviations can have poor coverage. Yang, Wang, and He (2016) proposed an adjustment to the posterior covariance matrix that produces asymptotically valid intervals. However, we show that this adjustment is sensitive to the choice of scale parameter for the AL likelihood and can lead to poor coverage when the sample size is small to moderate. We therefore propose using Infinitesimal Jackknife (IJ) standard errors (Giordano & Broderick, 2023). These standard errors do not require resampling but can be obtained from a single MCMC run. We also propose a version of IJ standard errors for clustered data. Simulations and applications to real data show that the IJ standard errors have good frequentist properties, both for independent and clustered data. We provide an R-package, IJSE, that computes IJ standard errors for clustered or independent data after estimation with the brms wrapper in R for Stan.

stat.ME↗

arXiv2024

Regularity of the free boundaries for the two-phase axisymmetric inviscid fluid

In the seminal paper (Alt, Caffarelli and Friedman, Trans. Amer. Math. Soc., 282, (1984).), the regularity of the free boundary of two-phase fluid in two dimensions via the so-called ACF energy functional was investigated. It was shown the $C^1$ regularity of the free boundaries and asserted that the two free boundaries coincide under some additional assumptions. Later on the standard technique of Harnack inequality could be applied to improve the regularity to $C^{1,η}$. A recent significant breakthrough in the regularity of two-phase fluid is due to De Philippis, Spolaor and Velichkov, who investigated the free boundary of the two-phase fluid with the two-phase functional (De Philippis, Spolaor and Velichkov, Invent. Math., 225, (2021).), and the $C^{1,η}$ regularity of the whole free boundaries was given in dimension two. Moreover, the free boundaries of the two-phase fluids do not coincide and the zero level set may process positive Lebesgue measure. In this paper, we consider the free boundaries for the two-phase axisymmetric fluid and show the free boundary is $C^{1,η}$ smooth. The Lebesgue measure of the zero level set of may also be positive, and the main difference lies in the degenerate elliptic operator and the free boundary conditions. More precisely, we use partial boundary Harnack inequalities and establish a linearized problem, whose regularity of the solutions implies the flatness decay of the two-phase free boundaries. Then the iteration argument gives the smoothness of the free boundaries.

math.AP↗

arXiv2024

Distributed-Order Fractional Graph Operating Network

We introduce the Distributed-order fRActional Graph Operating Network (DRAGON), a novel continuous Graph Neural Network (GNN) framework that incorporates distributed-order fractional calculus. Unlike traditional continuous GNNs that utilize integer-order or single fractional-order differential equations, DRAGON uses a learnable probability distribution over a range of real numbers for the derivative orders. By allowing a flexible and learnable superposition of multiple derivative orders, our framework captures complex graph feature updating dynamics beyond the reach of conventional models. We provide a comprehensive interpretation of our framework's capability to capture intricate dynamics through the lens of a non-Markovian graph random walk with node feature updating driven by an anomalous diffusion process over the graph. Furthermore, to highlight the versatility of the DRAGON framework, we conduct empirical evaluations across a range of graph learning tasks. The results consistently demonstrate superior performance when compared to traditional continuous GNN models. The implementation code is available at \url{https://github.com/zknus/NeurIPS-2024-DRAGON}.

cs.LG↗

arXiv2024

Edge centrality and the total variation of graph distributional signals

This short note is a supplement to [1], in which the total variation of graph distributional signals is introduced and studied. We introduce a different formulation of total variation and relate it to the notion of edge centrality. The relation provides a different perspective of total variation and may facilitate its computation.

eess.SP↗

arXiv2024

The Transferability of Downsamped Sparse Graph Convolutional Networks

To accelerate the training of graph convolutional networks (GCNs) on real-world large-scale sparse graphs, downsampling methods are commonly employed as a preprocessing step. However, the effects of graph sparsity and topological structure on the transferability of downsampling methods have not been rigorously analyzed or theoretically guaranteed, particularly when the topological structure is affected by graph sparsity. In this paper, we introduce a novel downsampling method based on a sparse random graph model and derive an expected upper bound for the transfer error. Our findings show that smaller original graph sizes, higher expected average degrees, and increased sampling rates contribute to reducing this upper bound. Experimental results validate the theoretical predictions. By incorporating both sparsity and topological similarity into the model, this study establishes an upper bound on the transfer error for downsampling in the training of large-scale sparse graphs and provides insight into the influence of topological structure on transfer performance.

cs.LG↗