arXiv subjects

Haoxiang Li

Publications and source records attributed to Haoxiang Li.

At least 55 records · Page 3Linked to original sources

arXiv2023

DDM-NET: End-to-end learning of keypoint feature Detection, Description and Matching for 3D localization

In this paper, we propose an end-to-end framework that jointly learns keypoint detection, descriptor representation and cross-frame matching for the task of image-based 3D localization. Prior art has tackled each of these components individually, purportedly aiming to alleviate difficulties in effectively train a holistic network. We design a self-supervised image warping correspondence loss for both feature detection and matching, a weakly-supervised epipolar constraints loss on relative camera pose learning, and a directional matching scheme that detects key-point features in a source image and performs coarse-to-fine correspondence search on the target image. We leverage this framework to enforce cycle consistency in our matching module. In addition, we propose a new loss to robustly handle both definite inlier/outlier matches and less-certain matches. The integration of these learning mechanisms enables end-to-end training of a single network performing all three localization components. Bench-marking our approach on public data-sets, exemplifies how such an end-to-end framework is able to yield more accurate localization that out-performs both traditional methods as well as state-of-the-art weakly supervised methods.

cs.CV↗

arXiv2023

Convergence Analysis of Data Augmentation Algorithms for Bayesian Robust Multivariate Linear Regression with Incomplete Data

Gaussian mixtures are commonly used for modeling heavy-tailed error distributions in robust linear regression. Combining the likelihood of a multivariate robust linear regression model with a standard improper prior distribution yields an analytically intractable posterior distribution that can be sampled using a data augmentation algorithm. When the response matrix has missing entries, there are unique challenges to the application and analysis of the convergence properties of the algorithm. Conditions for geometric ergodicity are provided when the incomplete data have a "monotone" structure. In the absence of a monotone structure, an intermediate imputation step is necessary for implementing the algorithm. In this case, we provide sufficient conditions for the algorithm to be Harris ergodic. Finally, we show that, when there is a monotone structure and intermediate imputation is unnecessary, intermediate imputation slows the convergence of the underlying Monte Carlo Markov chain, while post hoc imputation does not. An R package for the data augmentation algorithm is provided.

math.ST↗

arXiv2022

Boosted Dynamic Neural Networks

Early-exiting dynamic neural networks (EDNN), as one type of dynamic neural networks, has been widely studied recently. A typical EDNN has multiple prediction heads at different layers of the network backbone. During inference, the model will exit at either the last prediction head or an intermediate prediction head where the prediction confidence is higher than a predefined threshold. To optimize the model, these prediction heads together with the network backbone are trained on every batch of training data. This brings a train-test mismatch problem that all the prediction heads are optimized on all types of data in training phase while the deeper heads will only see difficult inputs in testing phase. Treating training and testing inputs differently at the two phases will cause the mismatch between training and testing data distributions. To mitigate this problem, we formulate an EDNN as an additive model inspired by gradient boosting, and propose multiple training techniques to optimize the model effectively. We name our method BoostNet. Our experiments show it achieves the state-of-the-art performance on CIFAR100 and ImageNet datasets in both anytime and budgeted-batch prediction modes. Our code is released at https://github.com/SHI-Labs/Boosted-Dynamic-Networks.

cs.LG↗

arXiv2022

Electronic structure and correlations in planar trilayer nickelate Pr4Ni3O8

The recent discovery of superconductivity in hole-doped planar nickelates R1-xSrNiO2 (R=Pr,Nd) raises the foundational question of how the electronic structure and electronic correlations of these Ni1+ compounds compare to those of the Cu2+ cuprate superconductors. Here, we present an Angle-Resolved Photoemission Spectroscopy (ARPES) study of the trilayer nickelate Pr4Ni3O8, revealing an electronic structure and Fermi surface very similar to that of the hole-doped cuprates but with a few critical differences. Specifically, the main portions of the Fermi surface are extremely similar to that of the bilayer cuprates, with an additional piece that can accommodate additional hole doping. We find that the electronic correlations are about twice as strong in the nickelates and are almost k-independent, indicating that they originate from a local effect-likely the Mott interaction, whereas the cuprate interactions are somewhat less local. Nevertheless, the nickelates still demonstrate an approximately linear in energy and linear in temperature scattering rate. Understanding the similarities and differences between these two related families of strongly-correlated novel superconductors is an important challenge.

cond-mat.supr-con↗

arXiv2022

Topological Electronic Structure Evolution with Symmetry Breaking Spin Reorientation in (Fe$_{1-x}$Co$_{x}$)Sn

Topological materials hosting kagome lattices have drawn considerable attention due to the interplay between topology, magnetism, and electronic correlations. The (Fe$_{1-x}$Co$_x$)Sn system not only hosts a kagome lattice but has a tunable symmetry breaking magnetic moment with temperature and doping. In this study, angle resolved photoemission spectroscopy and first principles calculations are used to investigate the interplay between the topological electronic structure and varying magnetic moment from the planar to axial antiferromagnetic phases. A theoretically predicted gap at the Dirac point is revealed in the low temperature axial phase but no gap opening is observed across a temperature dependent magnetic phase transition. However, topological surface bands are observed to shift in energy as the surface magnetic moment is reduced or becomes disordered over time during experimental measurements. The shifting surface bands may preclude the determination of a temperature dependent bulk gap but highlights the intricate connections between magnetism and topology with a surface/bulk dichotomy that can affect material properties and their interrogation.

cond-mat.str-el↗

arXiv2022

Phonon Chirality Induced by Vibronic-Orbital Coupling

The notion that phonons can carry pseudo-angular momentum has become popular in the last decade, with recent research efforts highlighting phonon chirality, Berry curvature of phonon band structure, and the phonon Hall effect. When a phonon is resonantly coupled to a crystal electric field excitation, a so-called vibronic bound state forms. Here, we observe angular momentum transfer of $δ$Jz = $\pm$1$\hbar$ between phonons and an orbital state in a vibronic bound state of a candidate quantum spin liquid. This observation has profound implications for the engineering of phonon band structure topology through chiral quasiparticle interactions.

cond-mat.mes-hall↗

arXiv2021

Realization of an intrinsic, ferromagnetic topological state in MnBi8Te13

The interplay between topology and magnetism is essential for realizing novel topological states including the axion insulator, the magnetic Weyl semimetal, etc. An intrinsically ferromagnetic topological material with only the topological bands at the charge neutrality energy has so far remained elusive. By rationally designing the natural heterostructure consisting of [MnBi2Te4] septuple layers and [Bi2Te3] quintuple layers, we report MnBi8Te13 as the first intrinsic ferromagnetic topological material with clean low-energy band structure. Based on the thermodynamic, transport and neutron diffraction measurements, our data show that despite the adjacent [MnBi2Te4] being 44.1 Å apart, MnBi8Te13 manifests long-range ferromagnetism below 10.5 K with strong coupling between magnetism and charge carriers. Our first-principles calculations and angle-resolved photoemission spectroscopy measurements further demonstrate that MnBi8Te13 is an intrinsic ferromagnetic axion state. Therefore, MnBi8Te13 serves as an ideal system to investigate rich emergent phenomena, including the quantized anomalous Hall effect and quantized topological magnetoelectric effect.

cond-mat.str-el↗

arXiv2021

Divergence of Majorana-Phonon Scattering in Kitaev Quantum Spin Liquid

Magnetoelastic interaction couples spin and lattice degrees of freedom and plays a key role in thermal transport properties of magnetic insulators. In the Kitaev quantum spin liquid, the low energy excitations are charge neutral Majorana fermions, which transform the magnetoelasctic interaction into Majorana-phonon scattering. Motivated by anomalous thermal properties of the Kitaev quantum spin liquid candidate RuCl$_3$, in this letter, we combine meV resolution inelastic x-ray scattering and theoretical calculation to examine the Majorana-phonon scattering. We analytically derive the velocity-dependent Majorana-phonon scattering and find a divergence when the acoustic phonons and the itinerant Majorana fermions have the same velocity. Based on the experimentally determined acoustic phonon velocity in RuCl$_3$, we estimate the range in the Kitaev interaction for which divergent Majorana-phonon scattering can happen. Our result opens a new avenue to uncover fractionalized quasiparticles in the Kitaev quantum spin liquid and emphasizes the critical role of lattice excitations in RuCl$_3$.

cond-mat.str-el↗

arXiv2021

Highly accelerated MR parametric mapping by undersampling the k-space and reducing the contrast number simultaneously with deep learning

Purpose: To propose a novel deep learning-based method called RG-Net (reconstruction and generation network) for highly accelerated MR parametric mapping by undersampling k-space and reducing the acquired contrast number simultaneously. Methods: The proposed framework consists of a reconstruction module and a generative module. The reconstruction module reconstructs MR images from the acquired few undersampled k-space data with the help of a data prior. The generative module then synthesizes the remaining multi-contrast images from the reconstructed images, where the exponential model is implicitly incorporated into the image generation through the supervision of fully sampled labels. The RG-Net was evaluated on the T1\r{ho} mapping data of knee and brain at different acceleration rates. Regional T1\r{ho} analysis for cartilage and the brain was performed to access the performance of RG-Net. Results: RG-Net yields a high-quality T1\r{ho} map at a high acceleration rate of 17. Compared with the competing methods that only undersample k-space, our framework achieves better performance in T1\r{ho} value analysis. Our method also improves quality of T1\r{ho} maps on patient with glioma. Conclusion: The proposed RG-Net that adopted a new strategy by undersampling k-space and reducing the contrast number simultaneously for fast MR parametric mapping, can achieve a high acceleration rate while maintaining good reconstruction quality. The generative module of our framework can also be used as an insert module in other fast MR parametric mapping methods. Keywords: Deep learning, convolutional neural network, fast MR parametric mapping

eess.IV↗

arXiv2021

Self-regulated growth of candidate topological superconducting parkerite by molecular beam epitaxy

Ternary chalcogenides such as the parkerites and shandites are a broad class of materials exhibiting rich diversity of transport and magnetic behavior as well as an array of topological phases including Weyl and Dirac nodes. However, they remain largely unexplored as high-quality epitaxial thin films. Here, we report the self-regulated growth of thin films of the strong spin-orbit coupled superconductor Pd3Bi2Se2 on SrTiO3 by molecular beam epitaxy. Films are found to grow in a self-regulated fashion, where, in excess Se, the temperature and relative flux ratio of Pd to Bi controls the formation of Pd3Bi2Se2 due to the combined volatility of Bi, Se, and Bi-Se bonded phases. The resulting films are shown to be of high structural quality, the stoichiometry is independent of the Pd:Bi and Se flux ratio and exhibit a superconducting transition temperature of 800 mK and critical field of 17.7 +/- 0.5 mT, as probed by transport as well as magnetometry. Understanding and navigating the growth of the chemically and structurally diverse classes of ternary chalcogenides opens a vast space for discovering new phenomena as well as enabling new applications.

cond-mat.mtrl-sci↗

arXiv2021

Spatial symmetry constraint of charge-ordered kagome superconductor CsV$_3$Sb$_5$

Elucidating the symmetry of intertwined orders in exotic superconductors is at the quantum frontier. Recent surface sensitive studies of the topological kagome superconductor CsV$_3$Sb$_5$ discovered a cascade 4a$_0$ superlattice below the charge density wave (CDW) ordering temperature, which can be related to the pair density modulations in the superconducting state. If the 4a$_0$ phase is a bulk and intrinsic property of the kagome lattice, this would form a striking analogy to the stripe order and pair density wave discovered in the cuprate high-temperature superconductors, and the cascade ordering found in twisted bilayer graphene. High-resolution X-ray diffraction has recently been established as an ultra-sensitive probe for bulk translational symmetry-breaking orders, even for short-range orders at the diffusive limit. Here, combining high-resolution X-ray diffraction, scanning tunneling microscopy and scanning transmission electron microscopy, we demonstrate that the 4a$_0$ superstructure emerges uniquely on the surface and hence exclude the 4a$_0$ phase as the origin of any bulk transport or spectroscopic anomaly. Crucially, we show that our detected 2$\times$2$\times$2 CDW order breaks the bulk rotational symmetry to C2, which can be the driver for the bulk nematic orders and nematic surface superlattices including the 4a$_0$ phase. Our high-resolution data impose decisive spatial symmetry constraints on emergent electronic orders in the kagome superconductor CsV$_3$Sb$_5$.

cond-mat.mtrl-sci↗

arXiv2021

Correlated Oxide Dirac Semimetal in the Extreme Quantum Limit

Quantum materials (QMs) with strong correlation and non-trivial topology are indispensable to next-generation information and computing technologies. Exploitation of topological band structure is an ideal starting point to realize correlated topological QMs. Herein, we report that strain-induced symmetry modification in correlated oxide SrNbO3 thin films creates an emerging topological band structure. Dirac electrons in strained SrNbO3 films reveal ultra-high mobility (100,000 cm2/Vs), exceptionally small effective mass (0.04me), and non-zero Berry phase. More importantly, strained SrNbO3 films reach the extreme quantum limit, exhibiting a sign of fractional occupation of Landau levels and giant mass enhancement. Our results suggest that symmetry-modified SrNbO3 is a rare example of a correlated topological QM, in which strong correlation of Dirac electrons leads to the realization of fractional occupation of Landau levels.

cond-mat.str-el↗

arXiv2021

Learning Dynamics via Graph Neural Networks for Human Pose Estimation and Tracking

Multi-person pose estimation and tracking serve as crucial steps for video understanding. Most state-of-the-art approaches rely on first estimating poses in each frame and only then implementing data association and refinement. Despite the promising results achieved, such a strategy is inevitably prone to missed detections especially in heavily-cluttered scenes, since this tracking-by-detection paradigm is, by nature, largely dependent on visual evidences that are absent in the case of occlusion. In this paper, we propose a novel online approach to learning the pose dynamics, which are independent of pose detections in current fame, and hence may serve as a robust estimation even in challenging scenarios including occlusion. Specifically, we derive this prediction of dynamics through a graph neural network~(GNN) that explicitly accounts for both spatial-temporal and visual information. It takes as input the historical pose tracklets and directly predicts the corresponding poses in the following frame for each tracklet. The predicted poses will then be aggregated with the detected poses, if any, at the same frame so as to produce the final pose, potentially recovering the occluded joints missed by the estimator. Experiments on PoseTrack 2017 and PoseTrack 2018 datasets demonstrate that the proposed method achieves results superior to the state of the art on both human pose estimation and tracking tasks.

cs.CV↗

arXiv2021

Breadcrumbs: Adversarial Class-Balanced Sampling for Long-tailed Recognition

The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered. While training with class-balanced sampling has been shown effective for this problem, it is known to over-fit to few-shot classes. It is hypothesized that this is due to the repeated sampling of examples and can be addressed by feature space augmentation. A new feature augmentation strategy, EMANATE, based on back-tracking of features across epochs during training, is proposed. It is shown that, unlike class-balanced sampling, this is an adversarial augmentation strategy. A new sampling procedure, Breadcrumb, is then introduced to implement adversarial class-balanced sampling without extra computation. Experiments on three popular long-tailed recognition datasets show that Breadcrumb training produces classifiers that outperform existing solutions to the problem.

cs.CV↗

arXiv2021

GistNet: a Geometric Structure Transfer Network for Long-Tailed Recognition

The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered. It is hypothesized that the well known tendency of standard classifier training to overfit to popular classes can be exploited for effective transfer learning. Rather than eliminating this overfitting, e.g. by adopting popular class-balanced sampling methods, the learning algorithm should instead leverage this overfitting to transfer geometric information from popular to low-shot classes. A new classifier architecture, GistNet, is proposed to support this goal, using constellations of classifier parameters to encode the class geometry. A new learning algorithm is then proposed for GeometrIc Structure Transfer (GIST), with resort to a combination of loss functions that combine class-balanced and random sampling to guarantee that, while overfitting to the popular classes is restricted to geometric parameters, it is leveraged to transfer class geometry from popular to few-shot classes. This enables better generalization for few-shot classes without the need for the manual specification of class weights, or even the explicit grouping of classes into different types. Experiments on two popular long-tailed recognition datasets show that GistNet outperforms existing solutions to this problem.

cs.CV↗

arXiv2021

Semi-supervised Long-tailed Recognition using Alternate Sampling

Main challenges in long-tailed recognition come from the imbalanced data distribution and sample scarcity in its tail classes. While techniques have been proposed to achieve a more balanced training loss and to improve tail classes data variations with synthesized samples, we resort to leverage readily available unlabeled data to boost recognition accuracy. The idea leads to a new recognition setting, namely semi-supervised long-tailed recognition. We argue this setting better resembles the real-world data collection and annotation process and hence can help close the gap to real-world scenarios. To address the semi-supervised long-tailed recognition problem, we present an alternate sampling framework combining the intuitions from successful methods in these two research areas. The classifier and feature embedding are learned separately and updated iteratively. The class-balanced sampling strategy has been implemented to train the classifier in a way not affected by the pseudo labels' quality on the unlabeled data. A consistency loss has been introduced to limit the impact from unlabeled data while leveraging them to update the feature embedding. We demonstrate significant accuracy improvements over other competitive methods on two datasets.

cs.CV↗

arXiv2021

Van der Waals Epitaxy on Freestanding Monolayer Graphene Membrane by MBE

Research on two-dimensional materials has expanded over the past two decades to become a central theme in condensed matter research today. Significant advances have been made in the synthesis and subsequent reassembly of these materials using mechanical methods into a vast array of hybrid structures with novel properties and ever-increasing potential applications. The key hurdles in realizing this potential are the challenges in controlling the atomic structure of these layered hybrid materials and the difficulties in harnessing their unique functionality with existing semiconductor nanofabrication techniques. Here we report on high-quality van der Waals epitaxial growth and characterization of a layered topological insulator on freestanding monolayer graphene transferred to different mechanical supports. This templated synthesis approach enables direct interrogation of interfacial atomic structure of these as-grown hybrid structures and opens a route towards creating device structures with more traditional semiconductor nanofabrication techniques.

cond-mat.mtrl-sci↗

arXiv2021

Beyond Visual Attractiveness: Physically Plausible Single Image HDR Reconstruction for Spherical Panoramas

HDR reconstruction is an important task in computer vision with many industrial needs. The traditional approaches merge multiple exposure shots to generate HDRs that correspond to the physical quantity of illuminance of the scene. However, the tedious capturing process makes such multi-shot approaches inconvenient in practice. In contrast, recent single-shot methods predict a visually appealing HDR from a single LDR image through deep learning. But it is not clear whether the previously mentioned physical properties would still hold, without training the network to explicitly model them. In this paper, we introduce the physical illuminance constraints to our single-shot HDR reconstruction framework, with a focus on spherical panoramas. By the proposed physical regularization, our method can generate HDRs which are not only visually appealing but also physically plausible. For evaluation, we collect a large dataset of LDR and HDR images with ground truth illuminance measures. Extensive experiments show that our HDR images not only maintain high visual quality but also top all baseline methods in illuminance prediction accuracy.

cs.CV↗