arXiv subjects

Lin Lu

Publications and source records attributed to Lin Lu.

52 records · Page 3Linked to original sources

arXiv2024

FedDriveScore: Federated Scoring Driving Behavior with a Mixture of Metric Distributions

Scoring the driving performance of various drivers on a unified scale, based on how safe or economical they drive on their daily trips, is essential for the driver profile task. Connected vehicles provide the opportunity to collect real-world driving data, which is advantageous for constructing scoring models. However, the lack of pre-labeled scores impede the use of supervised regression models and the data privacy issues hinder the way of traditionally data-centralized learning on the cloud side for model training. To address them, an unsupervised scoring method is presented without the need for labels while still preserving fairness and objectiveness compared to subjective scoring strategies. Subsequently, a federated learning framework based on vehicle-cloud collaboration is proposed as a privacy-friendly alternative to centralized learning. This framework includes a consistently federated version of the scoring method to reduce the performance degradation of the global scoring model caused by the statistical heterogeneous challenge of local data. Theoretical and experimental analysis demonstrate that our federated scoring model is consistent with the utility of the centrally learned counterpart and is effective in evaluating driving performance.

cs.LG↗

arXiv2024

Learning driving style embedding from GPS-derived moving patterns for driver identification

Learning fingerprint-like driving style representations is crucial to accurately identify who is behind the wheel in open driving situations. This study explores the learning of driving styles with GPS signals that are currently available in connected vehicles for short-term driver identification. First, an input driving trajectory is windowed into subtrajectories with fixed time lengths. Then, each subtrajectory is further divided into overlapping dynamic segments. For each segment, the local features are obtained by combining statistical and state transitional patterns. Finally, the driving style embedded in each subtrajectory is learned with the proposed regularized recurrent neural network (RNN) for short-term driver identification. We evaluate the impacts of key factors and the effectiveness of the proposed approach on the identification performance of 5 and 10 drivers. The results show that our proposed neural network structure, which complements movement statistics (MS) with state transitions (ST), provides better prediction performance than existing deep learning methods.

cs.CE↗

arXiv2023

Exploring the Robustness of Decentralized Training for Large Language Models

Decentralized training of large language models has emerged as an effective way to democratize this technology. However, the potential threats associated with this approach have not been carefully discussed, which would hinder the development of decentralized training infrastructures. This paper aims to initiate discussion towards this end by exploring the robustness of decentralized training from three main perspectives. First, we demonstrate the vulnerabilities inherent in decentralized training frameworks in terms of hardware, data, and models. Second, we highlight the fundamental difference between decentralized foundation model training and vanilla federated learning, where the security techniques employed in federated learning cannot be applied directly. Third, we discuss the essential components required for a robust and efficient decentralized training framework and present a case study by modeling a concrete threat model. Our objective in this vision paper is to emphasize the importance of addressing security concerns in the context of decentralized training for large language models.

cs.LG↗

arXiv2023

Learning Gradient Fields for Scalable and Generalizable Irregular Packing

The packing problem, also known as cutting or nesting, has diverse applications in logistics, manufacturing, layout design, and atlas generation. It involves arranging irregularly shaped pieces to minimize waste while avoiding overlap. Recent advances in machine learning, particularly reinforcement learning, have shown promise in addressing the packing problem. In this work, we delve deeper into a novel machine learning-based approach that formulates the packing problem as conditional generative modeling. To tackle the challenges of irregular packing, including object validity constraints and collision avoidance, our method employs the score-based diffusion model to learn a series of gradient fields. These gradient fields encode the correlations between constraint satisfaction and the spatial relationships of polygons, learned from teacher examples. During the testing phase, packing solutions are generated using a coarse-to-fine refinement mechanism guided by the learned gradient fields. To enhance packing feasibility and optimality, we introduce two key architectural designs: multi-scale feature extraction and coarse-to-fine relation extraction. We conduct experiments on two typical industrial packing domains, considering translations only. Empirically, our approach demonstrates spatial utilization rates comparable to, or even surpassing, those achieved by the teacher algorithm responsible for training data generation. Additionally, it exhibits some level of generalization to shape variations. We are hopeful that this method could pave the way for new possibilities in solving the packing problem.

cs.LG↗

arXiv2023

Two-stage MR Image Segmentation Method for Brain Tumors based on Attention Mechanism

Multimodal magnetic resonance imaging (MRI) can reveal different patterns of human tissue and is crucial for clinical diagnosis. However, limited by cost, noise and manual labeling, obtaining diverse and reliable multimodal MR images remains a challenge. For the same lesion, different MRI manifestations have great differences in background information, coarse positioning and fine structure. In order to obtain better generation and segmentation performance, a coordination-spatial attention generation adversarial network (CASP-GAN) based on the cycle-consistent generative adversarial network (CycleGAN) is proposed. The performance of the generator is optimized by introducing the Coordinate Attention (CA) module and the Spatial Attention (SA) module. The two modules can make full use of the captured location information, accurately locating the interested region, and enhancing the generator model network structure. The ability to extract the structure information and the detailed information of the original medical image can help generate the desired image with higher quality. There exist some problems in the original CycleGAN that the training time is long, the parameter amount is too large, and it is difficult to converge. In response to this problem, we introduce the Coordinate Attention (CA) module to replace the Res Block to reduce the number of parameters, and cooperate with the spatial information extraction network above to strengthen the information extraction ability. On the basis of CASP-GAN, an attentional generative cross-modality segmentation (AGCMS) method is further proposed. This method inputs the modalities generated by CASP-GAN and the real modalities into the segmentation network for brain tumor segmentation. Experimental results show that CASP-GAN outperforms CycleGAN and some state-of-the-art methods in PSNR, SSMI and RMSE in most tasks.

eess.IV↗

arXiv2023

Stochastic Porous Microstructures

Stochastic porous structures are ubiquitous in natural phenomena and have gained considerable traction across diverse domains owing to their exceptional physical properties. The recent surge in interest in microstructures can be attributed to their impressive attributes, such as a high strength-to-weight ratio, isotropic elasticity, and bio-inspired design principles. Notwithstanding, extant stochastic structures are predominantly generated via procedural modeling techniques, which present notable difficulties in representing geometric microstructures with periodic boundaries, thereby leading to intricate simulations and computational overhead. In this manuscript, we introduce an innovative method for designing stochastic microstructures that guarantees the periodicity of each microstructure unit to facilitate homogenization. We conceptualize each pore and the interconnecting tunnel between proximate pores as Gaussian kernels and leverage a modified version of the minimum spanning tree technique to assure pore connectivity. We harness the dart-throwing strategy to stochastically produce pore locations, tailoring the distribution law to enforce boundary periodicity. We subsequently employ the level-set technique to extract the stochastic microstructures. Conclusively, we adopt Wang tile rules to amplify the stochasticity at the boundary of the microstructure unit, concurrently preserving periodicity constraints among units. Our methodology offers facile parametric control of the designed stochastic microstructures. Experimental outcomes on 3D models manifest the superior isotropy and energy absorption performance of the stochastic porous microstructures. We further corroborate the efficacy of our modeling strategy through simulations of mechanical properties and empirical experiments.

cs.GR↗

arXiv2023

Generalized Pole-Residue Method for Dynamic Analysis of Nonlinear Systems based on Volterra Series

Dynamic systems characterized by second-order nonlinear ordinary differential equations appear in many fields of physics and engineering. To solve these kinds of problems, time-consuming step-by-step numerical integration methods and convolution methods based on Volterra series in the time domain have been widely used. In contrast, this work develops an efficient generalized pole-residue method based on the Volterra series performed in the Laplace domain. The proposed method involves two steps: (1) the Volterra kernels are decoupled in terms of Laguerre polynomials, and (2) the partial response related to a single Laguerre polynomial is obtained analytically in terms of the pole-residue method. Compared to the traditional pole-residue method for a linear system, one of the novelties of the pole-residue method in this paper is how to deal with the higher-order poles and their corresponding coefficients. Because the proposed method derives an explicit, continuous response function of time, it is much more efficient than traditional numerical methods. Unlike the traditional Laplace domain method, the proposed method is applicable to arbitrary irregular excitations. Because the natural response, forced response and cross response are naturally obtained in the solution procedure, meaningful mathematical and physical insights are gained. In numerical studies, systems with a known equation of motion and an unknown equation of motion are investigated. For each system, regular excitations and complex irregular excitations with different parameters are studied. Numerical studies validate the good accuracy and high efficiency of the proposed method by comparing it with the fourth-order Runge--Kutta method.

math.NA↗

arXiv2023

Automated deep reinforcement learning for real-time scheduling strategy of multi-energy system integrated with post-carbon and direct-air carbon captured system

The carbon-capturing process with the aid of CO2 removal technology (CDRT) has been recognised as an alternative and a prominent approach to deep decarbonisation. However, the main hindrance is the enormous energy demand and the economic implication of CDRT if not effectively managed. Hence, a novel deep reinforcement learning agent (DRL), integrated with an automated hyperparameter selection feature, is proposed in this study for the real-time scheduling of a multi-energy system coupled with CDRT. Post-carbon capture systems (PCCS) and direct-air capture systems (DACS) are considered CDRT. Various possible configurations are evaluated using real-time multi-energy data of a district in Arizona and CDRT parameters from manufacturers' catalogues and pilot project documentation. The simulation results validate that an optimised soft-actor critic (SAC) algorithm outperformed the TD3 algorithm due to its maximum entropy feature. We then trained four (4) SAC agents, equivalent to the number of considered case studies, using optimised hyperparameter values and deployed them in real time for evaluation. The results show that the proposed DRL agent can meet the prosumers' multi-energy demand and schedule the CDRT energy demand economically without specified constraints violation. Also, the proposed DRL agent outperformed rule-based scheduling by 23.65%. However, the configuration with PCCS and solid-sorbent DACS is considered the most suitable configuration with a high CO2 captured-released ratio of 38.54, low CO2 released indicator value of 2.53, and a 36.5% reduction in CDR cost due to waste heat utilisation and high absorption capacity of the selected sorbent. However, the adoption of CDRT is not economically viable at the current carbon price. Finally, we showed that CDRT would be attractive at a carbon price of 400-450USD/ton with the provision of tax incentives by the policymakers.

eess.SY↗

arXiv2022

PH-Net: Parallelepiped Microstructure Homogenization via 3D Convolutional Neural Networks

Microstructures are attracting academic and industrial interests with the rapid development of additive manufacturing. The numerical homogenization method has been well studied for analyzing mechanical behaviors of microstructures; however, it is too time-consuming to be applied to online computing or applications requiring high-frequency calling, e.g., topology optimization. Data-driven homogenization methods emerge as a more efficient choice but limit the microstructures into a cubic shape, which are infeasible to the periodic microstructures with a more general shape, e.g., parallelepiped. This paper introduces a fine-designed 3D convolutional neural network (CNN) for fast homogenization of parallel-shaped microstructures, named PH-Net. Superior to existing data-driven methods, PH-Net predicts the local displacements of microstructures under specified macroscope strains instead of direct homogeneous material, motivating us to present a label-free loss function based on minimal potential energy. For dataset construction, we introduce a shape-material transformation and voxel-material tensor to encode microstructure type,base material and boundary shape together as the input of PH-Net, such that it is CNN-friendly and enhances PH-Net on generalization in terms of microstructure type, base material, and boundary shape. PH-Net predicts homogenized properties with hundreds of acceleration compared to the numerical homogenization method and even supports online computing. Moreover, it does not require a labeled dataset and thus is much faster than current deep learning methods in training processing. Benefiting from predicting local displacement, PH-Net provides both homogeneous material properties and microscopic mechanical properties, e.g., strain and stress distribution, yield strength, etc. We design a group of physical experiments and verify the prediction accuracy of PH-Net.

cond-mat.mtrl-sci↗

arXiv2022

As-Continuous-As-Possible Extrusion Fabrication of Surface Models

We propose a novel computational framework for optimizing the toolpath continuity in fabricating surface models on an extrusion-based 3D printer. Toolpath continuity has been a critical issue for extrusion-based fabrications that affects both quality and efficiency. Transfer moves cause non-smoothor bumpy surfaces and get worse for materials with large inertia like clay. For surface models, the effects of continuity are even more severe, in terms of surface quality and model stability. In this paper, we introduce an original criterion "one-path-patch" (OPP), for representing a shell surface patch that can be traversed in one path considering fabrication constraints. We study the properties of an OPP and the merging operations for OPPs, and propose a bottom-up OPP merging procedure for decomposing the given shell surface into a minimal number of OPPs and generating the "as-continuous-as-possible" (ACAP) toolpath. Furthermore, we customize the path planning algorithm with a curved layer printing scheme, which reduces the staircase defect and improves the toolpath continuity via possibly connecting multiple segments. We evaluate the ACAP algorithm for both ceramic and thermoplastic materials, and results demonstrate that it improves the fabrication of surface models in both surface quality and efficiency.

cs.GR↗

arXiv2022

Study of a 43 day optical quasi-periodic oscillation for BL Lac S5 0716+714

The data for BL Lac object S5 0716+714 optical B, R and I bands are collected from November 10, 2017 to May 15, 2018. The raw data consists of 21396 quasi-simultaneous multi-band points, with 7132 data points for each band. To search for the quasi-periodic oscillation (QPO) signal of S5 0716+714, the Lomb Scargle periodogram (LSP) method and the Weighted Wavelet Z-transform (WWZ) method are used, and the light curve simulation method is utilized to estimate the significance level of the QPO. For the first time, a QPO with 43-day (> 95% confidence level) of approximate moon-like oscillations is revealed using these techniques. Furthermore, further research indicates that optical radiation is produced as a result of relativistic jet. This 43-day QPO can be explained by a spiral motion of a blob with velocity $\vec β$ along the jet. In addition, this 43-day QPO is superimposed on a longer QPO, and this superposition phenomenon can explain the linear upward trend. The viewing angle of the blob undergoes periodic changes, while jet precession results in a longer QPO. When the longer QPO fluxes are increasing state, a periodic rising tendency may be observed.

astro-ph.HE↗

arXiv2021

Online Updating Statistics for Heterogenous Updating Regressions via Homogenization Techniques

Under the environment of big data streams, it is a common situation where the variable set of a model may change according to the condition of data streams. In this paper, we propose a homogenization strategy to represent the heterogenous models that are gradually updated in the process of data streams. With the homogenized representations, we can easily construct various online updating statistics such as parameter estimation, residual sum of squares and $F$-statistic for the heterogenous updating regression models. The main difference from the classical scenarios is that the artificial covariates in the homogenized models are not identically distributed as the natural covariates in the original models, consequently, the related theoretical properties are distinct from the classical ones. The asymptotical properties of the online updating statistics are established, which show that the new method can achieve estimation efficiency and oracle property, without any constraint on the number of data batches. The behavior of the method is further illustrated by various numerical examples from simulation experiments.

stat.ME↗

arXiv2020

Fake Generated Painting Detection via Frequency Analysis

With the development of deep neural networks, digital fake paintings can be generated by various style transfer algorithms.To detect the fake generated paintings, we analyze the fake generated and real paintings in Fourier frequency domain and observe statistical differences and artifacts. Based on our observations, we propose Fake Generated Painting Detection via Frequency Analysis (FGPD-FA) by extracting three types of features in frequency domain. Besides, we also propose a digital fake painting detection database for assessing the proposed method. Experimental results demonstrate the excellence of the proposed method in different testing conditions.

cs.CV↗

arXiv2019

Prediction on Elastic Properties of Nb-doped Ni Systems

On the basis of the first principles simulation, the structure, formation enthalpy, and mechanical properties (elastic constant, bulk, and shear modulus and hardness) of five Nb-doped Ni systems are systematically studied. The calculated equilibrium volume increases with the Nb concentration increasing. The computational elastic constants and formation enthalpy indicate that all Nb-doped Ni systems are mechanically and thermodynamically stable in our research. The hardness of these systems also be predicted after the bulk modulus and shear modulus have been accurately calculated. The results show that the hardness increases with the Nb concentration increasing when the Nb concentration below 4.9%, beyond which the hardness will decrease within the scope of our study.

cond-mat.mtrl-sci↗

arXiv2015

Printed Perforated Lampshades for Continuous Projective Images

We present a technique for designing 3D-printed perforated lampshades, which project continuous grayscale images onto the surrounding walls. Given the geometry of the lampshade and a target grayscale image, our method computes a distribution of tiny holes over the shell, such that the combined footprints of the light emanating through the holes form the target image on a nearby diffuse surface. Our objective is to approximate the continuous tones and the spatial detail of the target image, to the extent possible within the constraints of the fabrication process. To ensure structural integrity, there are lower bounds on the thickness of the shell, the radii of the holes, and the minimal distances between adjacent holes. Thus, the holes are realized as thin tubes distributed over the lampshade surface. The amount of light passing through a single tube may be controlled by the tube's radius and by its direction (tilt angle). The core of our technique thus consists of determining a suitable configuration of the tubes: their distribution across the relevant portion of the lampshade, as well as the parameters (radius, tilt angle) of each tube. This is achieved by computing a capacity-constrained Voronoi tessellation over a suitably defined density function, and embedding a tube inside the maximal inscribed circle of each tessellation cell. The density function for a particular target image is derived from a series of simulated images, each corresponding to a different uniform density tube pattern on the lampshade.

cs.GR↗

arXiv2013

Assessing the Health of Richibucto Estuary with the Latent Health Factor Index

The ability to quantitatively assess the health of an ecosystem is often of great interest to those tasked with monitoring and conserving ecosystems. For decades, research in this area has relied upon multimetric indices of various forms. Although indices may be numbers, many are constructed based on procedures that are highly qualitative in nature, thus limiting the quantitative rigour of the practical interpretations made from these indices. The statistical modelling approach to construct the latent health factor index (LHFI) was recently developed to express ecological data, collected to construct conventional multimetric health indices, in a rigorous quantitative model that integrates qualitative features of ecosystem health and preconceived ecological relationships among such features. This hierarchical modelling approach allows (a) statistical inference of health for observed sites and (b) prediction of health for unobserved sites, all accompanied by formal uncertainty statements. Thus far, the LHFI approach has been demonstrated and validated on freshwater ecosystems. The goal of this paper is to adapt this approach to modelling estuarine ecosystem health, particularly that of the previously unassessed system in Richibucto in New Brunswick, Canada. Field data correspond to biotic health metrics that constitute the AZTI marine biotic index (AMBI) and abiotic predictors preconceived to influence biota. We also briefly discuss related LHFI research involving additional metrics that form the infaunal trophic index (ITI). Our paper is the first to construct a scientifically sensible model to rigorously identify the collective explanatory capacity of salinity, distance downstream, channel depth, and silt-clay content --- all regarded a priori as qualitatively important abiotic drivers --- towards site health in the Richibucto ecosystem.

stat.AP↗