Searcharxiv⌕ Search

arXiv subjects

Li Wang

Publications and source records attributed to Li Wang.

At least 289 records · Page 16Linked to original sources
arXiv2023

FMRT: Learning Accurate Feature Matching with Reconciliatory Transformer

Local Feature Matching, an essential component of several computer vision tasks (e.g., structure from motion and visual localization), has been effectively settled by Transformer-based methods. However, these methods only integrate long-range context information among keypoints with a fixed receptive field, which constrains the network from reconciling the importance of features with different receptive fields to realize complete image perception, hence limiting the matching accuracy. In addition, these methods utilize a conventional handcrafted encoding approach to integrate the positional information of keypoints into the visual descriptors, which limits the capability of the network to extract reliable positional encoding message. In this study, we propose Feature Matching with Reconciliatory Transformer (FMRT), a novel Transformer-based detector-free method that reconciles different features with multiple receptive fields adaptively and utilizes parallel networks to realize reliable positional encoding. Specifically, FMRT proposes a dedicated Reconciliatory Transformer (RecFormer) that consists of a Global Perception Attention Layer (GPAL) to extract visual descriptors with different receptive fields and integrate global context information under various scales, Perception Weight Layer (PWL) to measure the importance of various receptive fields adaptively, and Local Perception Feed-forward Network (LPFFN) to extract deep aggregated multi-scale local feature representation. Extensive experiments demonstrate that FMRT yields extraordinary performance on multiple benchmarks, including pose estimation, visual localization, homography estimation, and image matching.

cs.CV↗
arXiv2023

Design Considerations for Factorial Adaptive Multi-Arm Multi-Stage (FAST) Clinical Trials

Multi-Arm, Multi-Stage (MAMS) clinical trial designs allow for multiple therapies to be compared across a spectrum of clinical trial phases. MAMS designs can be categorized into several overarching design groups, including adaptive designs (AD) and multi-arm (MA) designs. Factorial clinical trials designs represent an additional group of designs which can provide increased efficiency relative to fixed, traditional designs. In this work, we explore design choices associated with Factorial Adaptive Multi-Arm Multi-Stage (FAST) designs, which represent the combination of factorial and MAMS designs. This category of trial can potentially offer benefits similar to both MAMS and factorial designs. This work is motivated by a proposed clinical trial under development.

stat.AP↗
arXiv2023

Audio compression-assisted feature extraction for voice replay attack detection

Replay attack is one of the most effective and simplest voice spoofing attacks. Detecting replay attacks is challenging, according to the Automatic Speaker Verification Spoofing and Countermeasures Challenge 2021 (ASVspoof 2021), because they involve a loudspeaker, a microphone, and acoustic conditions (e.g., background noise). One obstacle to detecting replay attacks is finding robust feature representations that reflect the channel noise information added to the replayed speech. This study proposes a feature extraction approach that uses audio compression for assistance. Audio compression compresses audio to preserve content and speaker information for transmission. The missed information after decompression is expected to contain content- and speaker-independent information (e.g., channel noise added during the replay process). We conducted a comprehensive experiment with a few data augmentation techniques and 3 classifiers on the ASVspoof 2021 physical access (PA) set and confirmed the effectiveness of the proposed feature extraction approach. To the best of our knowledge, the proposed approach achieves the lowest EER at 22.71% on the ASVspoof 2021 PA evaluation set.

cs.SD↗
arXiv2023

MonoGAE: Roadside Monocular 3D Object Detection with Ground-Aware Embeddings

Although the majority of recent autonomous driving systems concentrate on developing perception methods based on ego-vehicle sensors, there is an overlooked alternative approach that involves leveraging intelligent roadside cameras to help extend the ego-vehicle perception ability beyond the visual range. We discover that most existing monocular 3D object detectors rely on the ego-vehicle prior assumption that the optical axis of the camera is parallel to the ground. However, the roadside camera is installed on a pole with a pitched angle, which makes the existing methods not optimal for roadside scenes. In this paper, we introduce a novel framework for Roadside Monocular 3D object detection with ground-aware embeddings, named MonoGAE. Specifically, the ground plane is a stable and strong prior knowledge due to the fixed installation of cameras in roadside scenarios. In order to reduce the domain gap between the ground geometry information and high-dimensional image features, we employ a supervised training paradigm with a ground plane to predict high-dimensional ground-aware embeddings. These embeddings are subsequently integrated with image features through cross-attention mechanisms. Furthermore, to improve the detector's robustness to the divergences in cameras' installation poses, we replace the ground plane depth map with a novel pixel-level refined ground plane equation map. Our approach demonstrates a substantial performance advantage over all previous monocular 3D object detectors on widely recognized 3D detection benchmarks for roadside cameras. The code and pre-trained models will be released soon.

cs.CV↗
arXiv2023

BEVHeight++: Toward Robust Visual Centric 3D Object Detection

While most recent autonomous driving system focuses on developing perception methods on ego-vehicle sensors, people tend to overlook an alternative approach to leverage intelligent roadside cameras to extend the perception ability beyond the visual range. We discover that the state-of-the-art vision-centric bird's eye view detection methods have inferior performances on roadside cameras. This is because these methods mainly focus on recovering the depth regarding the camera center, where the depth difference between the car and the ground quickly shrinks while the distance increases. In this paper, we propose a simple yet effective approach, dubbed BEVHeight++, to address this issue. In essence, we regress the height to the ground to achieve a distance-agnostic formulation to ease the optimization process of camera-only perception methods. By incorporating both height and depth encoding techniques, we achieve a more accurate and robust projection from 2D to BEV spaces. On popular 3D detection benchmarks of roadside cameras, our method surpasses all previous vision-centric methods by a significant margin. In terms of the ego-vehicle scenario, our BEVHeight++ possesses superior over depth-only methods. Specifically, it yields a notable improvement of +1.9% NDS and +1.1% mAP over BEVDepth when evaluated on the nuScenes validation set. Moreover, on the nuScenes test set, our method achieves substantial advancements, with an increase of +2.8% NDS and +1.7% mAP, respectively.

cs.CV↗
arXiv2023

A Stochastic Online Forecast-and-Optimize Framework for Real-Time Energy Dispatch in Virtual Power Plants under Uncertainty

Aggregating distributed energy resources in power systems significantly increases uncertainties, in particular caused by the fluctuation of renewable energy generation. This issue has driven the necessity of widely exploiting advanced predictive control techniques under uncertainty to ensure long-term economics and decarbonization. In this paper, we propose a real-time uncertainty-aware energy dispatch framework, which is composed of two key elements: (i) A hybrid forecast-and-optimize sequential task, integrating deep learning-based forecasting and stochastic optimization, where these two stages are connected by the uncertainty estimation at multiple temporal resolutions; (ii) An efficient online data augmentation scheme, jointly involving model pre-training and online fine-tuning stages. In this way, the proposed framework is capable to rapidly adapt to the real-time data distribution, as well as to target on uncertainties caused by data drift, model discrepancy and environment perturbations in the control process, and finally to realize an optimal and robust dispatch solution. The proposed framework won the championship in CityLearn Challenge 2022, which provided an influential opportunity to investigate the potential of AI application in the energy domain. In addition, comprehensive experiments are conducted to interpret its effectiveness in the real-life scenario of smart building energy management.

eess.SY↗
arXiv2023

The role of tidal interactions in the formation of slowly rotating early-type stars in young star clusters

The split main sequences found in the colour-magnitude diagrams of star clusters younger than ~600 Myr are suggested to be caused by the dichotomy of stellar rotation rates of upper main-sequence stars. Tidal interactions have been suggested as a possible explanation of the dichotomy of the stellar rotation rates. This hypothesis proposes that the slow rotation rates of stars along the split main sequences are caused by tidal interactions in binaries. To test this scenario, we measured the variations in the radial velocities of slowly rotating stars along the split main sequence of the young Galactic cluster NGC 2422 (~90 Myr) using spectra obtained at multiple epochs with the Canada-France-Hawai'i Telescope. Our results show that most slowly rotating stars are not radial-velocity variables. Using the theory of dynamical tides, we find that the binary separations necessary to fully or partially synchronise our spectroscopic targets, on time-scales shorter than the cluster age, predict much larger radial velocity variations across multiple-epoch observations, or a much larger radial velocity dispersion at a single epoch, than the observed values. This indicates that tidal interactions are not the dominant mechanism to form slowly rotating stars along the split main sequences. As the observations of the rotation velocity distribution among B- and A-type stars in binaries of larger separations hint at a much stronger effect of braking with age, we discuss the consequences of relaxing the constraints of the dynamical tides theory.

astro-ph.SR↗
arXiv2023

Deep Learning and Inverse Problems

Machine Learning (ML) methods and tools have gained great success in many data, signal, image and video processing tasks, such as classification, clustering, object detection, semantic segmentation, language processing, Human-Machine interface, etc. In computer vision, image and video processing, these methods are mainly based on Neural Networks (NN) and in particular Convolutional NN (CNN), and more generally Deep NN. Inverse problems arise anywhere we have indirect measurement. As, in general, those inverse problems are ill-posed, to obtain satisfactory solutions for them needs prior information. Different regularization methods have been proposed, where the problem becomes the optimization of a criterion with a likelihood term and a regularization term. The main difficulty, however, in great dimensional real applications, remains the computational cost. Using NN, and in particular Deep Learning (DL) surrogate models and approximate computation, can become very helpful. In this work, we focus on NN and DL particularly adapted for inverse problems. We consider two cases: First the case where the forward operator is known and used as physics constraint, the second more general data driven DL methods.

cs.LG↗
arXiv2023

Deep Learning and Bayesian inference for Inverse Problems

Inverse problems arise anywhere we have indirect measurement. As, in general they are ill-posed, to obtain satisfactory solutions for them needs prior knowledge. Classically, different regularization methods and Bayesian inference based methods have been proposed. As these methods need a great number of forward and backward computations, they become costly in computation, in particular, when the forward or generative models are complex and the evaluation of the likelihood becomes very costly. Using Deep Neural Network surrogate models and approximate computation can become very helpful. However, accounting for the uncertainties, we need first understand the Bayesian Deep Learning and then, we can see how we can use them for inverse problems. In this work, we focus on NN, DL and more specifically the Bayesian DL particularly adapted for inverse problems. We first give details of Bayesian DL approximate computations with exponential families, then we will see how we can use them for inverse problems. We consider two cases: First the case where the forward operator is known and used as physics constraint, the second more general data driven DL methods. keyword: Neural Network, Variational Bayesian inference, Bayesian Deep Learning (DL), Inverse problems, Physics based DL.

stat.ML↗
arXiv2023

Engineering mobility in quasiperiodic lattices with exact mobility edges

We investigate the effect of an additional modulation parameter $δ$ on the mobility properties of quasiperiodic lattices described by a generalized Ganeshan-Pixley-Das Sarma model with two on site modulation parameters. For the case with bounded quasiperiodic potential, we unveil the existence of self-duality relation, independent of $δ$. By applying Avila's global theory, we analytically derive Lyapunov exponents in the whole parameter space, which enables us to determine mobility edges or anomalous mobility edges exactly. Our analytical results indicate that the mobility edge equation is described by two curves and their intersection with the spectrum gives the true mobility edge. Tuning the strength parameter $δ$ can change the spectrum of the quasiperiodic lattice, and thus engineers the mobility of quasi-periodic systems, giving rise to completely extended, partially localized, and completely localized regions. For the case with unbounded quasiperiodic potential, we also obtain the analytical expression of the anomalous mobility edge, which separates localized states from critical states. By increasing the strength parameter $δ$, we find that the critical states can be destroyed gradually and finally vanishes.

cond-mat.dis-nn↗
arXiv2023

Observation of the Breakdown of Optical Phonon Splitting in a Two-dimensional Polar Monolayer

Phonon splitting of the longitudinal optical and transverse optical modes (LO-TO splitting), a ubiquitous phenomenon in three-dimensional (3D) polar materials, is essential for the formation of the 3D phonon polaritons. Theories predict that the LO-TO splitting will break down in two-dimensional (2D) polar systems, but direct experimental verification is still missing. Here, using monolayer hexagonal boron nitride (h-BN) as a prototypical example, we report the direct observation of the breakdown of LO-TO splitting and the finite slope of the LO phonons at the center of the Brillouin zone in 2D polar materials by inelastic electron scattering spectroscopy. Interestingly, the slope of the LO phonon in our measurements is lower than the theoretically predicted value for a freestanding monolayer due to the screening of the Cu foil substrate. This enables the phonon polaritons (PhPs) in monolayer h-BN/Cu foil to exhibit ultra-slow group velocity (~ 5 x 10^-6 c, c is the speed of light) and ultra-high confinement (~ 4000 times smaller wavelength than that of light). Our work reveals the universal law of the LO phonons in 2D polar materials and lays a physical foundation for future research on 2D PhPs.

cond-mat.mtrl-sci↗
arXiv2023

Brain Anatomy Prior Modeling to Forecast Clinical Progression of Cognitive Impairment with Structural MRI

Brain structural MRI has been widely used to assess the future progression of cognitive impairment (CI). Previous learning-based studies usually suffer from the issue of small-sized labeled training data, while there exist a huge amount of structural MRIs in large-scale public databases. Intuitively, brain anatomical structures derived from these public MRIs (even without task-specific label information) can be used to boost CI progression trajectory prediction. However, previous studies seldom take advantage of such brain anatomy prior. To this end, this paper proposes a brain anatomy prior modeling (BAPM) framework to forecast the clinical progression of cognitive impairment with small-sized target MRIs by exploring anatomical brain structures. Specifically, the BAPM consists of a pretext model and a downstream model, with a shared brain anatomy-guided encoder to model brain anatomy prior explicitly. Besides the encoder, the pretext model also contains two decoders for two auxiliary tasks (i.e., MRI reconstruction and brain tissue segmentation), while the downstream model relies on a predictor for classification. The brain anatomy-guided encoder is pre-trained with the pretext model on 9,344 auxiliary MRIs without diagnostic labels for anatomy prior modeling. With this encoder frozen, the downstream model is then fine-tuned on limited target MRIs for prediction. We validate the BAPM on two CI-related studies with T1-weighted MRIs from 448 subjects. Experimental results suggest the effectiveness of BAPM in (1) four CI progression prediction tasks, (2) MR image reconstruction, and (3) brain tissue segmentation, compared with several state-of-the-art methods.

eess.IV↗
arXiv2023

Testing for intrinsic multifractality in the global grain spot market indices: A multifractal detrended fluctuation analysis

Grains account for more than 50% of the calories consumed by people worldwide, and military conflicts, pandemics, climate change, and soaring grain prices all have vital impacts on food security. However, the complex price behavior of the global grain spot markets has not been well understood. A recent study performed multifractal moving average analysis (MF-DMA) of the Grains & Oilseeds Index (GOI) and its sub-indices of wheat, maize, soyabeans, rice, and barley and found that only the maize and barley sub-indices exhibit an intrinsic multifractal nature with convincing evidence. Here, we utilize multifractal fluctuation analysis (MF-DFA) to investigate the same problem. Extensive statistical tests confirm the presence of intrinsic multifractality in the maize and barley sub-indices and the absence of intrinsic multifractality in the wheat and rice sub-indices. Different from the MF-DMA results, the MF-DFA results suggest that there is also intrinsic multifractality in the GOI and soyabeans sub-indices. Our comparative analysis does not provide conclusive information about the GOI and soyabeans and highlights the high complexity of the global grain spot markets.

q-fin.ST↗
arXiv2023

Momentum spirals in multiphoton pair production revisited

Spirals in multiphoton pair production are revisited by two counter-rotating fields with time delay for different cycles in pulse. Novel findings include that for subcycle fields, the remarkable spiral structure in the momentum spectrum can be still caused by a large time delay compared to the previous study for supercycle case where it is easier to be generated by a small time delay. And also there exist a range of critical polarization values for the spirals appearance corresponding to the different cycle number. The relative phase difference between two fields causes not only severe symmetry breaking of the momentum spectra pattern and spiral, but also a significant change for the shape and the number of spiral arm. Upon the number density, it is found a more sensitive to the cycle number, in particularly, it is enhanced by more than one order of magnitude for small cycle pulse, while it is increased about few times when the time delay is small. These results provide an abundant theoretical testbed for the possible experimental observation on the multiphoton pair production in future. Meanwhile, it is applicable to regard the particles momentum signatures as a new probing to the laser field information with it from the vacuum.

hep-ph↗
arXiv2023

Suppressing Instability in a Vlasov-Poisson System by an External Electric Field Through Constrained Optimization

Fusion energy offers the potential for the generation of clean, safe, and nearly inexhaustible energy. While notable progress has been made in recent years, significant challenges persist in achieving net energy gain. Improving plasma confinement and stability stands as a crucial task in this regard and requires optimization and control of the plasma system. In this work, we deploy a PDE-constrained optimization formulation that uses a kinetic description for plasma dynamics as the constraint. This is to optimize, over all possible controllable external electric fields, the stability of the plasma dynamics under the condition that the Vlasov--Poisson (VP) equation is satisfied. For computing the functional derivative with respect to the external field in the optimization updates, the adjoint equation is derived. Furthermore, in the discrete setting, where we employ the semi-Lagrangian method as the forward solver, we also explicitly formulate the corresponding adjoint solver and the gradient as the discrete analogy to the adjoint equation and the Frechet derivative. A distinct feature we observed of this constrained optimization is the complex landscape of the objective function and the existence of numerous local minima, largely due to the hyperbolic nature of the VP system. To overcome this issue, we utilize a gradient-accelerated genetic algorithm, leveraging the advantages of the genetic algorithm's exploration feature to cover a broader search of the solution space and the fast local convergence aided by the gradient information. We show that our algorithm obtains good electric fields that are able to maintain a prescribed profile in a beam shaping problem and uses nonlinear effects to suppress plasma instability in a two-stream configuration.

math.NA↗
arXiv2023

An immersed boundary method for the fluid--structure--thermal interaction in rarefied gas flow

An immersed boundary method for the fluid--structure--thermal interaction in rarefied gas flow is presented. In this method, the slip model is incorporated with the penalty immersed boundary method to address the velocity and temperature jump conditions at the fluid--structure interface in rarefied gas flow within slip regime. In this method, the compressible flow governed by Navier-Stokes equations are solved by using high-order finite difference method; the elastic solid is solved by using finite element method; the fluid and solid are solved independently and the fluid--structure--thermal interaction are achieved by using a penalty method in a partitioned way. Several validations are conducted including Poiseuille flow in a 2D pipe, flow around a 2D NACA airfoil, moving square cylinder in a 2D pipe, flow around a sphere and moving sphere in quiescent flow. The numerical results from present method show good agreement with the previous published data obtained by other methods, and it confirms the the good ability of the proposed method in handling fluid--structure--thermal interaction for both weakly compressible and highly compressible rarefied gas flow. To overcome the incapability of Navier-Stokes equations at high local Knudsen numbers in supersonic flow, an artificial viscosity is introduced to ease the sharp transition at the shock wave front. Inspired by Martian exploration, the application of proposed method to study the aerodynamics of flapping wing in rarefied gas flow is conducted in both 2D and 3D domains, to obtain some insights for the flapping-wing aerial vehicles operating in Martian environment.

physics.flu-dyn↗
arXiv2023

Option pricing under jump diffusion model

We provide an European option pricing formula written in the form of an infinite series of Black Scholes type terms under double Levy jumps model, where both the interest rate and underlying price are driven by Levy process. The series solution converges with a radius of convergence, and it is complemented by some numerical experiments to demonstrate its speed of convergence.

q-fin.PR↗
arXiv2023

Connection Sensitivity Matters for Training-free DARTS: From Architecture-Level Scoring to Operation-Level Sensitivity Analysis

The recently proposed training-free NAS methods abandon the training phase and design various zero-cost proxies as scores to identify excellent architectures, arousing extreme computational efficiency for neural architecture search. In this paper, we raise an interesting problem: can we properly measure the operation importance in DARTS through a training-free way, with avoiding the parameter-intensive bias? We investigate this question through the lens of edge connectivity, and provide an affirmative answer by defining a connectivity concept, ZERo-cost Operation Sensitivity (ZEROS), to score the importance of candidate operations in DARTS at initialization. By devising an iterative and data-agnostic manner in utilizing ZEROS for NAS, our novel trial leads to a framework called training free differentiable architecture search (FreeDARTS). Based on the theory of Neural Tangent Kernel (NTK), we show the proposed connectivity score provably negatively correlated with the generalization bound of DARTS supernet after convergence under gradient descent training. In addition, we theoretically explain how ZEROS implicitly avoids parameter-intensive bias in selecting architectures, and empirically show the searched architectures by FreeDARTS are of comparable size. Extensive experiments have been conducted on a series of search spaces, and results have demonstrated that FreeDARTS is a reliable and efficient baseline for neural architecture search.

cs.LG↗