arXiv subjects

Guan Wang

Publications and source records attributed to Guan Wang.

At least 55 records · Page 3Linked to original sources

arXiv2025

TRKT: Weakly Supervised Dynamic Scene Graph Generation with Temporal-enhanced Relation-aware Knowledge Transferring

Dynamic Scene Graph Generation (DSGG) aims to create a scene graph for each video frame by detecting objects and predicting their relationships. Weakly Supervised DSGG (WS-DSGG) reduces annotation workload by using an unlocalized scene graph from a single frame per video for training. Existing WS-DSGG methods depend on an off-the-shelf external object detector to generate pseudo labels for subsequent DSGG training. However, detectors trained on static, object-centric images struggle in dynamic, relation-aware scenarios required for DSGG, leading to inaccurate localization and low-confidence proposals. To address the challenges posed by external object detectors in WS-DSGG, we propose a Temporal-enhanced Relation-aware Knowledge Transferring (TRKT) method, which leverages knowledge to enhance detection in relation-aware dynamic scenarios. TRKT is built on two key components:(1)Relation-aware knowledge mining: we first employ object and relation class decoders that generate category-specific attention maps to highlight both object regions and interactive areas. Then we propose an Inter-frame Attention Augmentation strategy that exploits optical flow for neighboring frames to enhance the attention maps, making them motion-aware and robust to motion blur. This step yields relation- and motion-aware knowledge mining for WS-DSGG. (2) we introduce a Dual-stream Fusion Module that integrates category-specific attention maps into external detections to refine object localization and boost confidence scores for object proposals. Extensive experiments demonstrate that TRKT achieves state-of-the-art performance on Action Genome dataset. Our code is avaliable at https://github.com/XZPKU/TRKT.git.

cs.CV↗

arXiv2025

Post-estimation Adjustments in Data-driven Decision-making with Applications in Pricing

The predict-then-optimize (PTO) framework is a standard approach in data-driven decision-making, where a decision-maker first estimates an unknown parameter from historical data and then uses this estimate to solve an optimization problem. While widely used for its simplicity and modularity, PTO can lead to suboptimal decisions because the estimation step does not account for the structure of the downstream optimization problem. We study a class of problems where the objective function, evaluated at the PTO decision, is asymmetric with respect to estimation errors. This asymmetry causes the expected outcome to be systematically degraded by noise in the parameter estimate, as the penalty for underestimation differs from that of overestimation. To address this, we develop a data-driven post-estimation adjustment that improves decision quality while preserving the practicality and modularity of PTO. We show that when the objective function satisfies a particular curvature condition, based on the ratio of its third and second derivatives, the adjustment simplifies to a closed-form expression. This condition holds for a broad range of pricing problems, including those with linear, log-linear, and power-law demand models. Under this condition, we establish theoretical guarantees that our adjustment uniformly and asymptotically outperforms standard PTO, and we precisely characterize the resulting improvement. Additionally, we extend our framework to multi-parameter optimization and settings with biased estimators. Numerical experiments demonstrate that our method consistently improves revenue, particularly in small-sample regimes where estimation uncertainty is most pronounced. This makes our approach especially well-suited for pricing new products or in settings with limited historical price variation.

math.OC↗

arXiv2025

GBGC: Efficient and Adaptive Graph Coarsening via Granular-ball Computing

The objective of graph coarsening is to generate smaller, more manageable graphs while preserving key information of the original graph. Previous work were mainly based on the perspective of spectrum-preserving, using some predefined coarsening rules to make the eigenvalues of the Laplacian matrix of the original graph and the coarsened graph match as much as possible. However, they largely overlooked the fact that the original graph is composed of subregions at different levels of granularity, where highly connected and similar nodes should be more inclined to be aggregated together as nodes in the coarsened graph. By combining the multi-granularity characteristics of the graph structure, we can generate coarsened graph at the optimal granularity. To this end, inspired by the application of granular-ball computing in multi-granularity, we propose a new multi-granularity, efficient, and adaptive coarsening method via granular-ball (GBGC), which significantly improves the coarsening results and efficiency. Specifically, GBGC introduces an adaptive granular-ball graph refinement mechanism, which adaptively splits the original graph from coarse to fine into granular-balls of different sizes and optimal granularity, and constructs the coarsened graph using these granular-balls as supernodes. In addition, compared with other state-of-the-art graph coarsening methods, the processing speed of this method can be increased by tens to hundreds of times and has lower time complexity. The accuracy of GBGC is almost always higher than that of the original graph due to the good robustness and generalization of the granular-ball computing, so it has the potential to become a standard graph data preprocessing method.

cs.AI↗

arXiv2025

Capturing Aperiodic Temporal Dynamics of EEG Signals through Stochastic Fluctuation Modeling

Electrophysiological brain signals, such as electroencephalography (EEG), exhibit both periodic and aperiodic components, with the latter often modeled as 1/f noise and considered critical to cognitive and neurological processes. Although various theoretical frameworks have been proposed to account for aperiodic activity, its scale-invariant and long-range temporal dependency remain insufficiently explained. Drawing on neural fluctuation theory, we propose a novel framework that parameterizes intrinsic stochastic neural fluctuations to account for aperiodic dynamics. Within this framework, we introduce two key parameters-self-similarity and scale factor-to characterize these fluctuations. Our findings reveal that EEG fluctuations exhibit self-similar and non-stable statistical properties, challenging the assumptions of conventional stochastic models in neural dynamical modeling. Furthermore, the proposed parameters enable the reconstruction of EEG-like signals that faithfully replicate the aperiodic spectrum, including the characteristic 1/f spectral profile, and long range dependency. By linking structured neural fluctuations to empirically observed aperiodic EEG activity, this work offers deeper mechanistic insights into brain dynamics, resulting in a more robust biomarker candidate than the traditional 1/f slope, and provides a computational methodology for generating biologically plausible neurophysiological signals.

q-bio.NC↗

arXiv2025

GBO:AMulti-Granularity Optimization Algorithm via Granular-ball for Continuous Problems

Optimization problems aim to find the optimal solution, which is becoming increasingly complex and difficult to solve. Traditional evolutionary optimization methods always overlook the granular characteristics of solution space. In the real scenario of numerous optimizations, the solution space is typically partitioned into sub-regions characterized by varying degree distributions. These sub-regions present different granularity characteristics at search potential and difficulty. Considering the granular characteristics of the solution space, the number of coarse-grained regions is smaller than the number of points, so the calculation is more efficient. On the other hand, coarse-grained characteristics are not easily affected by fine-grained sample points, so the calculation is more robust. To this end, this paper proposes a new multi-granularity evolutionary optimization method, namely the Granular-ball Optimization (GBO) algorithm, which characterizes and searches the solution space from coarse to fine. Specifically, using granular-balls instead of traditional points for optimization increases the diversity and robustness of the random search process. At the same time, the search range in different iteration processes is limited by the radius of granular-balls, covering the solution space from large to small. The mechanism of granular-ball splitting is applied to continuously split and evolve the large granular-balls into smaller ones for refining the solution space. Extensive experiments on commonly used benchmarks have shown that GBO outperforms popular and advanced evolutionary algorithms. The code can be found in the supporting materials.

cs.LG↗

arXiv2025

A Token-FCM based risk assessment method for complex engineering designs

Engineering design risks could cause unaffordable losses, and thus risk assessment plays a critical role in engineering design. On the other hand, the high complexity of modern engineering designs makes it difficult to assess risks effectively and accurately due to the complex two-way, dynamic causal-effect risk relations in engineering designs. To address this problem, this paper proposes a new risk assessment method called token fuzzy cognitive map (Token-FCM). Its basic idea is to model the two-way causal-risk relations with the FCM method, and then augment FCM with a token mechanism to model the dynamics in causal-effect risk relations. Furthermore, the fuzzy sets and the group decision-making method are introduced to initialize the Token-FCM method so that comprehensive and accurate risk assessments can be attained. The effectiveness of the proposed method has been demonstrated by a real example of engine design for a horizontal directional drilling machine.

cs.CE↗

arXiv2024

Are Expressive Models Truly Necessary for Offline RL?

Among various branches of offline reinforcement learning (RL) methods, goal-conditioned supervised learning (GCSL) has gained increasing popularity as it formulates the offline RL problem as a sequential modeling task, therefore bypassing the notoriously difficult credit assignment challenge of value learning in conventional RL paradigm. Sequential modeling, however, requires capturing accurate dynamics across long horizons in trajectory data to ensure reasonable policy performance. To meet this requirement, leveraging large, expressive models has become a popular choice in recent literature, which, however, comes at the cost of significantly increased computation and inference latency. Contradictory yet promising, we reveal that lightweight models as simple as shallow 2-layer MLPs, can also enjoy accurate dynamics consistency and significantly reduced sequential modeling errors against large expressive models by adopting a simple recursive planning scheme: recursively planning coarse-grained future sub-goals based on current and target information, and then executes the action with a goal-conditioned policy learned from data rela-beled with these sub-goal ground truths. We term our method Recursive Skip-Step Planning (RSP). Simple yet effective, RSP enjoys great efficiency improvements thanks to its lightweight structure, and substantially outperforms existing methods, reaching new SOTA performances on the D4RL benchmark, especially in multi-stage long-horizon tasks.

cs.LG↗

arXiv2024

PO-VINS: An Efficient and Robust Pose-Only Visual-Inertial State Estimator With LiDAR Enhancement

The pose adjustment (PA) with a pose-only visual representation has been proven equivalent to the bundle adjustment (BA), while significantly improving the computational efficiency. However, the pose-only solution has not yet been properly considered in a tightly-coupled visual-inertial state estimator (VISE) with a normal configuration for real-time navigation. In this study, we propose a tightly-coupled LiDAR-enhanced VISE, named PO-VINS, with a full pose-only form for visual and LiDAR-depth measurements. Based on the pose-only visual representation, we derive the analytical depth uncertainty, which is then employed for rejecting LiDAR depth outliers. Besides, we propose a multi-state constraint (MSC)-based LiDAR-depth measurement model with a pose-only form, to balance efficiency and robustness. The pose-only visual and LiDAR-depth measurements and the IMU-preintegration measurements are tightly integrated under the factor graph optimization framework to perform efficient and accurate state estimation. Exhaustive experimental results on private and public datasets indicate that the proposed PO-VINS yields improved or comparable accuracy to sate-of-the-art methods. Compared to the baseline method LE-VINS, the state-estimation efficiency of PO-VINS is improved by 33% and 56% on the laptop PC and the onboard ARM computer, respectively. Besides, PO-VINS yields higher accuracy and robustness than LE-VINS by employing the proposed uncertainty-based outlier-culling method and the MSC-based measurement model for LiDAR depth.

cs.RO↗

arXiv2024

A Riemannian Approach for Spatiotemporal Analysis and Generation of 4D Tree-shaped Structures

We propose the first comprehensive approach for modeling and analyzing the spatiotemporal shape variability in tree-like 4D objects, i.e., 3D objects whose shapes bend, stretch, and change in their branching structure over time as they deform, grow, and interact with their environment. Our key contribution is the representation of tree-like 3D shapes using Square Root Velocity Function Trees (SRVFT). By solving the spatial registration in the SRVFT space, which is equipped with an L2 metric, 4D tree-shaped structures become time-parameterized trajectories in this space. This reduces the problem of modeling and analyzing 4D tree-like shapes to that of modeling and analyzing elastic trajectories in the SRVFT space, where elasticity refers to time warping. In this paper, we propose a novel mathematical representation of the shape space of such trajectories, a Riemannian metric on that space, and computational tools for fast and accurate spatiotemporal registration and geodesics computation between 4D tree-shaped structures. Leveraging these building blocks, we develop a full framework for modelling the spatiotemporal variability using statistical models and generating novel 4D tree-like structures from a set of exemplars. We demonstrate and validate the proposed framework using real 4D plant data.

cs.CV↗

arXiv2024

Crafting Efficient Fine-Tuning Strategies for Large Language Models

This paper addresses the challenges of efficiently fine-tuning large language models (LLMs) by exploring data efficiency and hyperparameter optimization. We investigate the minimum data required for effective fine-tuning and propose a novel hyperparameter optimization method that leverages early-stage model performance. Our experiments demonstrate that fine-tuning with as few as 200 samples can improve model accuracy from 70\% to 88\% in a product attribute extraction task. We identify a saturation point of approximately 6,500 samples, beyond which additional data yields diminishing returns. Our proposed bayesian hyperparameter optimization method, which evaluates models at 20\% of total training time, correlates strongly with final model performance, with 4 out of 5 top early-stage models remaining in the top 5 at completion. This approach led to a 2\% improvement in accuracy over baseline models when evaluated on an independent test set. These findings offer actionable insights for practitioners, potentially reducing computational load and dependency on extensive datasets while enhancing overall performance of fine-tuned LLMs.

cs.CL↗

arXiv2024

OED: Towards One-stage End-to-End Dynamic Scene Graph Generation

Dynamic Scene Graph Generation (DSGG) focuses on identifying visual relationships within the spatial-temporal domain of videos. Conventional approaches often employ multi-stage pipelines, which typically consist of object detection, temporal association, and multi-relation classification. However, these methods exhibit inherent limitations due to the separation of multiple stages, and independent optimization of these sub-problems may yield sub-optimal solutions. To remedy these limitations, we propose a one-stage end-to-end framework, termed OED, which streamlines the DSGG pipeline. This framework reformulates the task as a set prediction problem and leverages pair-wise features to represent each subject-object pair within the scene graph. Moreover, another challenge of DSGG is capturing temporal dependencies, we introduce a Progressively Refined Module (PRM) for aggregating temporal context without the constraints of additional trackers or handcrafted trajectories, enabling end-to-end optimization of the network. Extensive experiments conducted on the Action Genome benchmark demonstrate the effectiveness of our design. The code and models are available at \url{https://github.com/guanw-pku/OED}.

cs.CV↗

arXiv2024

OpenChat: Advancing Open-source Language Models with Mixed-Quality Data

Nowadays, open-source large language models like LLaMA have emerged. Recent developments have incorporated supervised fine-tuning (SFT) and reinforcement learning fine-tuning (RLFT) to align these models with human goals. However, SFT methods treat all training data with mixed quality equally, while RLFT methods require high-quality pairwise or ranking-based preference data. In this study, we present a novel framework, named OpenChat, to advance open-source language models with mixed-quality data. Specifically, we consider the general SFT training data, consisting of a small amount of expert data mixed with a large proportion of sub-optimal data, without any preference labels. We propose the C(onditioned)-RLFT, which regards different data sources as coarse-grained reward labels and learns a class-conditioned policy to leverage complementary data quality information. Interestingly, the optimal policy in C-RLFT can be easily solved through single-stage, RL-free supervised learning, which is lightweight and avoids costly human preference labeling. Through extensive experiments on three standard benchmarks, our openchat-13b fine-tuned with C-RLFT achieves the highest average performance among all 13b open-source language models. Moreover, we use AGIEval to validate the model generalization performance, in which only openchat-13b surpasses the base model. Finally, we conduct a series of analyses to shed light on the effectiveness and robustness of OpenChat. Our code, data, and models are publicly available at https://github.com/imoneoi/openchat and https://huggingface.co/openchat.

cs.CL↗

arXiv2024

A Tampering Risk of Fiber-Based Frequency Synchronization Networks and Its Countermeasures

Fiber optic networks are used worldwide and have been regarded as excellent media for transmitting time-frequency (TF) signals. In the past decades, fiber-based TF synchronization techniques have been extensively studied. Instruments based on these techniques have been successfully applied. With the increasing application of TF synchronization instruments, their security has become an important issue. Unfortunately, the security risks of fiber-based frequency synchronization (FbFS) instruments have been overlooked. This paper proposes a frequency tampering method called "frequency lens". On a 200 km fiber link, we demonstrate a frequency tampering scenario using a frequency lens-enabled frequency tampering module (FTM). On the user side, the frequency value of the recovered 100 MHz signal can be stealthily altered within a range of 100 MHz-100 Hz to 100 MHz+100 Hz, while the frequency dissemination stability of the system remains normal. Related to this tampering risk, potential hazards in three different application scenarios, which rely on precise frequency references, are analyzed. Two countermeasures are also proposed to solve this tampering risk.

physics.ins-det↗

arXiv2024

Detecting misinformation through Framing Theory: the Frame Element-based Model

In this paper, we delve into the rapidly evolving challenge of misinformation detection, with a specific focus on the nuanced manipulation of narrative frames - an under-explored area within the AI community. The potential for Generative AI models to generate misleading narratives underscores the urgency of this problem. Drawing from communication and framing theories, we posit that the presentation or 'framing' of accurate information can dramatically alter its interpretation, potentially leading to misinformation. We highlight this issue through real-world examples, demonstrating how shifts in narrative frames can transmute fact-based information into misinformation. To tackle this challenge, we propose an innovative approach leveraging the power of pre-trained Large Language Models and deep neural networks to detect misinformation originating from accurate facts portrayed under different frames. These advanced AI techniques offer unprecedented capabilities in identifying complex patterns within unstructured data critical for examining the subtleties of narrative frames. The objective of this paper is to bridge a significant research gap in the AI domain, providing valuable insights and methodologies for tackling framing-induced misinformation, thus contributing to the advancement of responsible and trustworthy AI technologies. Several experiments are intensively conducted and experimental results explicitly demonstrate the various impact of elements of framing theory proving the rationale of applying framing theory to increase the performance in misinformation detection.

cs.CL↗

arXiv2023

Elastic Shape Analysis of Tree-like 3D Objects using Extended SRVF Representation

How can one analyze detailed 3D biological objects, such as neurons and botanical trees, that exhibit complex geometrical and topological variation? In this paper, we develop a novel mathematical framework for representing, comparing, and computing geodesic deformations between the shapes of such tree-like 3D objects. A hierarchical organization of subtrees characterizes these objects -- each subtree has the main branch with some side branches attached -- and one needs to match these structures across objects for meaningful comparisons. We propose a novel representation that extends the Square-Root Velocity Function (SRVF), initially developed for Euclidean curves, to tree-shaped 3D objects. We then define a new metric that quantifies the bending, stretching, and branch sliding needed to deform one tree-shaped object into the other. Compared to the current metrics, such as the Quotient Euclidean Distance (QED) and the Tree Edit Distance (TED), the proposed representation and metric capture the full elasticity of the branches (i.e., bending and stretching) as well as the topological variations (i.e., branch death/birth and sliding). It completely avoids the shrinkage that results from the edge collapse and node split operations of the QED and TED metrics. We demonstrate the utility of this framework in comparing, matching, and computing geodesics between biological objects such as neurons and botanical trees. The framework is also applied to various shape analysis tasks: (i) symmetry analysis and symmetrization of tree-shaped 3D objects, (ii) computing summary statistics (means and modes of variations) of populations of tree-shaped 3D objects, (iii) fitting parametric probability distributions to such populations, and (iv) finally synthesizing novel tree-shaped 3D objects through random sampling from estimated probability distributions.

cs.LG↗

arXiv2023

Training A Multi-stage Deep Classifier with Feedback Signals

Multi-Stage Classifier (MSC) - several classifiers working sequentially in an arranged order and classification decision is partially made at each step - is widely used in industrial applications for various resource limitation reasons. The classifiers of a multi-stage process are usually Neural Network (NN) models trained independently or in their inference order without considering the signals from the latter stages. Aimed at two-stage binary classification process, the most common type of MSC, we propose a novel training framework, named Feedback Training. The classifiers are trained in an order reverse to their actual working order, and the classifier at the later stage is used to guide the training of initial-stage classifier via a sample weighting method. We experimentally show the efficacy of our proposed approach, and its great superiority under the scenario of few-shot training.

cs.LG↗

arXiv2023

An underground fiber cable discrimination method based on laser interferometer

The maintenance and repair of optical fiber networks often requires the discrimination of underground cables among a group of buried ones. Presently, the methods commonly used are either inefficient or harmful to the fiber cable. In this paper, we evaluate the effectiveness of fiber optic vibration sensing method on underground fiber cable discrimination. We find that the typical vibration sensing method-distributed acoustic sensing (DAS) is not suitable for fiber cable discrimination. Especially for long distance scenario, due to "frequency grafting" effect, the DAS method will give a false response on the spectrum of knock events during cable discrimination process. In response, we propose an underground fiber cable discrimination method based on laser interferometer, and demonstrate it on the 40-km urban fiber cable connecting Tsinghua University and Yongding Road. It can give an obvious and real response to the knock event, and can be used in practical applications.

physics.ins-det↗

arXiv2023

Evolving Connectivity for Recurrent Spiking Neural Networks

Recurrent spiking neural networks (RSNNs) hold great potential for advancing artificial general intelligence, as they draw inspiration from the biological nervous system and show promise in modeling complex dynamics. However, the widely-used surrogate gradient-based training methods for RSNNs are inherently inaccurate and unfriendly to neuromorphic hardware. To address these limitations, we propose the evolving connectivity (EC) framework, an inference-only method for training RSNNs. The EC framework reformulates weight-tuning as a search into parameterized connection probability distributions, and employs Natural Evolution Strategies (NES) for optimizing these distributions. Our EC framework circumvents the need for gradients and features hardware-friendly characteristics, including sparse boolean connections and high scalability. We evaluate EC on a series of standard robotic locomotion tasks, where it achieves comparable performance with deep neural networks and outperforms gradient-trained RSNNs, even solving the complex 17-DoF humanoid task. Additionally, the EC framework demonstrates a two to three fold speedup in efficiency compared to directly evolving parameters. By providing a performant and hardware-friendly alternative, the EC framework lays the groundwork for further energy-efficient applications of RSNNs and advances the development of neuromorphic devices.

cs.NE↗