arXiv subjects

Le Zhang

Publications and source records attributed to Le Zhang.

At least 217 records · Page 12Linked to original sources

arXiv2021

Detection of cross-correlation between CMB Lensing and low-density points

Low Density Points (LDPs, \citet{2019ApJ...874....7D}), obtained by removing high-density regions of observed galaxies, can trace the Large-Scale Structures (LSSs) of the universe. In particular, it offers an intriguing opportunity to detect weak gravitational lensing from low-density regions. In this work, we investigate tomographic cross-correlation between Planck CMB lensing maps and LDP-traced LSSs, where LDPs are constructed from the DR8 data release of the DESI legacy imaging survey, with about $10^6$-$10^7$ galaxies. We find that, due to the large sky coverage (20,000 deg$^2$) and large redshift depth ($z\leq 1.2$), a significant detection ($10σ$--$30σ$) of the CMB lensing-LDP cross-correlation in all six redshift bins can be achieved, with a total significance of $\sim 53σ$ over $ \ell\le1024$. Moreover, the measurements are in good agreement with a theoretical template constructed from our numerical simulation in the WMAP 9-year $Λ$CDM cosmology. A scaling factor for the lensing amplitude $A_{\rm lens}$ is constrained to $A_{\rm lens}=1\pm0.12$ for $z<0.2$, $A_{\rm lens}=1.07\pm0.07$ for $0.2<z<0.4$ and $A_{\rm lens}=1.07\pm0.05$ for $0.4<z<0.6$, with the r-band absolute magnitude cut of $-21.5$ for LDP selection. A variety of tests have been performed to check the detection reliability, against variations in LDP samples and galaxy magnitude cuts, masks, CMB lensing maps, multipole $\ell$ cuts, sky regions, and photo-z bias. We also perform a cross-correlation measurement between CMB lensing and galaxy number density, which is consistent with the CMB lensing-LDP cross-correlation. This work therefore further convincingly demonstrates that LDP is a competitive tracer of LSS.

astro-ph.CO↗

arXiv2021

On Automatic Data Augmentation for 3D Point Cloud Classification

Data augmentation is an important technique to reduce overfitting and improve learning performance, but existing works on data augmentation for 3D point cloud data are based on heuristics. In this work, we instead propose to automatically learn a data augmentation strategy using bilevel optimization. An augmentor is designed in a similar fashion to a conditional generator and is optimized by minimizing a base model's loss on a validation set when the augmented input is used for training the model. This formulation provides a more principled way to learn data augmentation on 3D point clouds. We evaluate our approach on standard point cloud classification tasks and a more challenging setting with pose misalignment between training and validation/test sets. The proposed strategy achieves competitive performance on both tasks and we provide further insight into the augmentor's ability to learn the validation set distribution.

cs.CV↗

arXiv2021

Semantic Edge Detection with Diverse Deep Supervision

Semantic edge detection (SED), which aims at jointly extracting edges as well as their category information, has far-reaching applications in domains such as semantic segmentation, object proposal generation, and object recognition. SED naturally requires achieving two distinct supervision targets: locating fine detailed edges and identifying high-level semantics. Our motivation comes from the hypothesis that such distinct targets prevent state-of-the-art SED methods from effectively using deep supervision to improve results. To this end, we propose a novel fully convolutional neural network using diverse deep supervision (DDS) within a multi-task framework where bottom layers aim at generating category-agnostic edges, while top layers are responsible for the detection of category-aware semantic edges. To overcome the hypothesized supervision challenge, a novel information converter unit is introduced, whose effectiveness has been extensively evaluated on SBD and Cityscapes datasets.

cs.CV↗

arXiv2021

Free Lunch for Co-Saliency Detection: Context Adjustment

We unveil a long-standing problem in the prevailing co-saliency detection systems: there is indeed inconsistency between training and testing. Constructing a high-quality co-saliency detection dataset involves time-consuming and labor-intensive pixel-level labeling, which has forced most recent works to rely instead on semantic segmentation or saliency detection datasets for training. However, the lack of proper co-saliency and the absence of multiple foreground objects in these datasets can lead to spurious variations and inherent biases learned by models. To tackle this, we introduce the idea of counterfactual training through context adjustment and propose a "cost-free" group-cut-paste (GCP) procedure to leverage off-the-shelf images and synthesize new samples. Following GCP, we collect a novel dataset called Context Adjustment Training (CAT). CAT consists of 33,500 images, which is four times larger than the current co-saliency detection datasets. All samples are automatically annotated with high-quality mask annotations, object categories, and edge maps. Extensive experiments on recent benchmarks are conducted, show that CAT can improve various state-of-the-art models by a large margin (5% ~ 25%). We hope that the scale, diversity, and quality of our dataset can benefit researchers in this area and beyond. Our dataset will be publicly accessible through our project page.

cs.CV↗

arXiv2021

Fast scalar quadratic maximum likelihood estimators for the CMB B-mode power spectrum

Constructing a fast and efficient estimator for the B-mode power spectrum of cosmic microwave background (CMB) is of critical importance for CMB science. For a general CMB survey, the Quadratic Maximum Likelihood (QML) estimator for CMB polarization has been proved to be the optimal estimator with minimal uncertainties, but it is computationally very expensive. In this article, we propose two new QML methods for B-mode power spectrum estimation. We use the Smith-Zaldarriaga approach to prepare pure-B mode map, and E-mode recycling method to obtain a leakage free B-mode map. We then use the scalar QML estimator to analyze the scalar pure-B map (QML-SZ) or B-mode map (QML-TC). The QML-SZ and QML-TC estimators have similar error bars as the standard QML estimators but their computational cost is nearly one order of magnitude smaller. The basic idea is that one can construct the pure B-mode CMB map by using the E-B separation method proposed by Smith-Zaldarriaga (SZ) or the one considering the template cleaning (TC) technique, then apply QML estimator to these scalar fields. By simulating potential observations of space-based and ground-based detectors, we test the reliability of these estimators by comparing them with the corresponding results of the traditional QML estimator and the pure B-mode pseudo-Cl estimator.

astro-ph.CO↗

arXiv2021

Unsupervised Scale-consistent Depth Learning from Video

We propose a monocular depth estimator SC-Depth, which requires only unlabelled videos for training and enables the scale-consistent prediction at inference time. Our contributions include: (i) we propose a geometry consistency loss, which penalizes the inconsistency of predicted depths between adjacent views; (ii) we propose a self-discovered mask to automatically localize moving objects that violate the underlying static scene assumption and cause noisy signals during training; (iii) we demonstrate the efficacy of each component with a detailed ablation study and show high-quality depth estimation results in both KITTI and NYUv2 datasets. Moreover, thanks to the capability of scale-consistent prediction, we show that our monocular-trained deep networks are readily integrated into the ORB-SLAM2 system for more robust and accurate tracking. The proposed hybrid Pseudo-RGBD SLAM shows compelling results in KITTI, and it generalizes well to the KAIST dataset without additional training. Finally, we provide several demos for qualitative evaluation.

cs.CV↗

arXiv2021

Dual Universality and Unconventional Phase Diagram of a Clean 2D Superconductor with Tunable Disorders

Weakly disordered two-dimensional (2D) superconductors can host richer quantum phase transitions than their highly-disordered counterparts. This is due to the insertion of a metallic state in the transition between a superconductor and an insulator. Disorders were predicted to show profound influences on the boundaries surrounding this intermediate metallic phase, affecting the existence of a metallic ground state called Bose metal and the dynamic processes in a quantum Griffiths state. Here we present a study on quantum phase transitions of a clean 2D superconductor, MoS2, under a perpendicular magnetic field as a function of disorder strength that is tuned electrostatically. We found two universal scaling behaviors independent of disorders: a power-law scaling applies for all metallic states, and an activated dynamical scaling characterizes transitions between quantum Griffiths state and weakly localized metal. A study over a range of disorder states in this unexplored clear regime suggests that a Bose metal ground state is expected only after reaching a critical cleanness. Whereas, stronger disorder stabilizes a true 2D superconductivity and enhances the quantum Griffiths phase.

cond-mat.supr-con↗

arXiv2021

Reconstructing simulated CMB polarization power spectra with the Analytical Blind Separation method

Multi-frequency observations are needed to separate the CMB from foregrounds and accurately extract cosmological information from the data. The Analytical Blind Separation (ABS) method is dedicated to extracting the CMB power spectrum from multi-frequency observations in the presence of contamination from astrophysical foreground emission and instrumental noise. In this study, we apply the ABS method to simulated sky maps as could be observed with a future space-borne survey, in order to test the method's capability for determining the CMB polarization $E$- and $B$-mode power spectra. We present the ABS method performance on simulations for both a full-sky analysis and for an analysis concentrating on sky regions less impacted by Galactic foreground emission. We discuss the origin and minimization of biases in the estimated CMB polarization angular power spectra. We find that the ABS method performs quite well for the analysis of full-sky observations at intermediate and small angular scales, in spite of strong foreground contamination. On the largest scales, extra work is still required to reduce biases of various origins and the impact of confusion between CMB $E$ and $B$ polarization for partial-sky analyses.

astro-ph.CO↗

arXiv2021

Regularized Densely-connected Pyramid Network for Salient Instance Segmentation

Much of the recent efforts on salient object detection (SOD) have been devoted to producing accurate saliency maps without being aware of their instance labels. To this end, we propose a new pipeline for end-to-end salient instance segmentation (SIS) that predicts a class-agnostic mask for each detected salient instance. To better use the rich feature hierarchies in deep networks and enhance the side predictions, we propose the regularized dense connections, which attentively promote informative features and suppress non-informative ones from all feature pyramids. A novel multi-level RoIAlign based decoder is introduced to adaptively aggregate multi-level features for better mask predictions. Such strategies can be well-encapsulated into the Mask R-CNN pipeline. Extensive experiments on popular benchmarks demonstrate that our design significantly outperforms existing \sArt competitors by 6.3\% (58.6\% vs. 52.3\%) in terms of the AP metric.The code is available at https://github.com/yuhuan-wu/RDPNet.

cs.CV↗

arXiv2021

The design of the Ali CMB Polarization Telescope receiver

Ali CMB Polarization Telescope (AliCPT-1) is the first CMB degree-scale polarimeter to be deployed on the Tibetan plateau at 5,250m above sea level. AliCPT-1 is a 90/150 GHz 72 cm aperture, two-lens refracting telescope cooled down to 4 K. Alumina lenses, 800mm in diameter, image the CMB in a 33.4° field of view on a 636mm wide focal plane. The modularized focal plane consists of dichroic polarization-sensitive Transition-Edge Sensors (TESes). Each module includes 1,704 optically active TESes fabricated on a 150mm diameter silicon wafer. Each TES array is read out with a microwave multiplexing readout system capable of a multiplexing factor up to 2,048. Such a large multiplexing factor has allowed the practical deployment of tens of thousands of detectors, enabling the design of a receiver that can operate up to 19 TES arrays for a total of 32,376 TESes. AliCPT-1 leverages the technological advancements in the detector design from multiple generations of previously successful feedhorn-coupled polarimeters, and in the instrument design from BICEP-3, but applied on a larger scale. The cryostat receiver is currently under integration and testing. During the first deployment year, the focal plane will be populated with up to 4 TES arrays. Further TES arrays will be deployed in the following years, fully populating the focal plane with 19 arrays on the fourth deployment year. Here we present the AliCPT-1 receiver design, and how the design has been optimized to meet the experimental requirements.

astro-ph.IM↗

arXiv2020

HIR4: Cosmological signatures imprinted on the cross correlation between 21-cm map and galaxy clustering

We explore the cosmological multitracer synergies between an emission line galaxy distribution from the Dark Energy Spectroscopic Instrument and a Tianlai Project 21-cm intensity map. We use simulated maps generated from a particle simulation in the light-cone volume (Horizon Run 4), sky-trimmed and including the effects of foreground contamination, its removal and instrument noise. We first validate how the foreground residual affects the recovered 21-cm signal by putting different levels of foreground contamination into the 21-cm maps. We find that the contamination cannot be ignored in the angular autocorrelation power spectra of HI even when it is small, but has no influence on the accuracy of the angular cross-power spectra between HI and galaxies. In the foreground-cleaned map case, as information is lost in the cleaning procedure, there is also a bias in the cross-correlation power spectrum. However, we found that the bias from the cross-correlation power spectrum is scale-independent, which is easily parameterized as part of the model, while the offset in the HI autocorrelation power spectrum is non-linear. In particular, we tested that the cross-correlation power also benefits from the cancellation of the bias in the power spectrum measurement that is induced by the instrument noise, which changes the shape of the autocorrelation power spectra but leaves the cross-correlation power unaffected. We then modelled the angular cross-correlation power spectra to fit the baryon acoustic oscillation feature in the broad-band shape of the angular cross-correlation power spectrum, including contamination from the residual foreground and the effect of instrument noise. We forecast a constraint on the angular diameter distance $D_\mathrm{A}$ for the Tianlai Pathfinder redshift $0.775<z<1.03$, giving a distance measurement with a precision of 2.7\% at that redshift.

astro-ph.CO↗

arXiv2020

Disentangling Human Error from the Ground Truth in Segmentation of Medical Images

Recent years have seen increasing use of supervised learning methods for segmentation tasks. However, the predictive performance of these algorithms depends on the quality of labels. This problem is particularly pertinent in the medical image domain, where both the annotation cost and inter-observer variability are high. In a typical label acquisition process, different human experts provide their estimates of the "true" segmentation labels under the influence of their own biases and competence levels. Treating these noisy labels blindly as the ground truth limits the performance that automatic segmentation algorithms can achieve. In this work, we present a method for jointly learning, from purely noisy observations alone, the reliability of individual annotators and the true segmentation label distributions, using two coupled CNNs. The separation of the two is achieved by encouraging the estimated annotators to be maximally unreliable while achieving high fidelity with the noisy training data. We first define a toy segmentation dataset based on MNIST and study the properties of the proposed algorithm. We then demonstrate the utility of the method on three public medical imaging segmentation datasets with simulated (when necessary) and real diverse annotations: 1) MSLSC (multiple-sclerosis lesions); 2) BraTS (brain tumours); 3) LIDC-IDRI (lung abnormalities). In all cases, our method outperforms competing methods and relevant baselines particularly in cases where the number of annotations is small and the amount of disagreement is large. The experiments also show strong ability to capture the complex spatial characteristics of annotators' mistakes.

cs.CV↗

arXiv2020

Generalized Zero-Shot Learning via VAE-Conditioned Generative Flow

Generalized zero-shot learning (GZSL) aims to recognize both seen and unseen classes by transferring knowledge from semantic descriptions to visual representations. Recent generative methods formulate GZSL as a missing data problem, which mainly adopts GANs or VAEs to generate visual features for unseen classes. However, GANs often suffer from instability, and VAEs can only optimize the lower bound on the log-likelihood of observed data. To overcome the above limitations, we resort to generative flows, a family of generative models with the advantage of accurate likelihood estimation. More specifically, we propose a conditional version of generative flows for GZSL, i.e., VAE-Conditioned Generative Flow (VAE-cFlow). By using VAE, the semantic descriptions are firstly encoded into tractable latent distributions, conditioned on that the generative flow optimizes the exact log-likelihood of the observed visual features. We ensure the conditional latent distribution to be both semantic meaningful and inter-class discriminative by i) adopting the VAE reconstruction objective, ii) releasing the zero-mean constraint in VAE posterior regularization, and iii) adding a classification regularization on the latent variables. Our method achieves state-of-the-art GZSL results on five well-known benchmark datasets, especially for the significant improvement in the large-scale setting. Code is released at https://github.com/guyuchao/VAE-cFlow-ZSL.

cs.CV↗

arXiv2020

HIR4: cosmology from a simulated neutral hydrogen full sky using Horizon Run 4

The distribution of cosmological neutral hydrogen will provide a new window into the large-scale structure of the Universe with the next generation of radio telescopes and surveys. The observation of this material, through 21cm line emission, will be confused by foreground emission in the same frequencies. Even after these foregrounds are removed, the reconstructed map may not exactly match the original cosmological signal, which will introduce systematic errors and offset into the measured correlations. In this paper, we simulate future surveys of neutral hydrogen using the Horizon Run 4 (HR4) cosmological N-body simulation. We generate HI intensity maps from the HR4 halo catalogue, and combine with foreground radio emission maps from the Global Sky Model, to create accurate simulations over the entire sky. We simulate the HI sky for the frequency range 700-800 MHz, matching the sensitivity of the Tianlai pathfinder. We test the accuracy of the fastICA, PCA and log-polynomial fitting foreground removal methods to recover the input cosmological angular power spectrum and measure the parameters. We show the effect of survey noise levels and beam sizes on the recovered the cosmological constraints. We find that while the reconstruction removes power from the cosmological 21cm distribution on large-scales, we can correct for this and recover the input parameters in the noise-free case. However, the effect of noise and beam size of the Tianlai pathfinder prevents accurate recovery of the cosmological parameters when using only intensity mapping information.

astro-ph.CO↗

arXiv2020

Fuzzy SLIC: Fuzzy Simple Linear Iterative Clustering

Most superpixel methods are sensitive to noise and cannot control the superpixel number precisely. To solve these problems, in this paper, we propose a robust superpixel method called fuzzy simple linear iterative clustering (Fuzzy SLIC), which adopts a local spatial fuzzy C-means clustering and dynamic fuzzy superpixels. We develop a fast and precise superpixel number control algorithm called onion peeling (OP) algorithm. Fuzzy SLIC is insensitive to most types of noise, including Gaussian, salt and pepper, and multiplicative noise. The OP algorithm can control the superpixel number accurately without reducing much computational efficiency. In the validation experiments, we tested the Fuzzy SLIC and OP algorithm and compared them with state-of-the-art methods on the BSD500 and Pascal VOC2007 benchmarks. The experiment results show that our methods outperform state-of-the-art techniques in both noise-free and noisy environments.

eess.IV↗

arXiv2020

Valley-Hall-like second-order photonic topological insulators in Kagome lattice

Valley-Hall-like second-order photonic topological insulators are designed in Kagome-lattice photonic crystals with C3v point-group symmetry. The photonic crystal consists of circular air holes in pure dielectric materials. Different from conventional valley-Hall photonic topological insulators characterized by valley Chern numbers, the proposed insulators have topological invariants described by quantized electric polarization. Topological transition can be realized by tuning the structural size and topological edge states appear at the interface between photonic crystals with different topological phases, preserving important features of valley-Hall photonic insulators such as valley transport with little backscattering. The proposed photonic crystal also support zero-dimensional corner states in oblique corners, showing its second-order topological insulator signature. This work presents the possibility to realize topologically protected reflection suppressed waveguides and local cavities in the same platform.

physics.optics↗

arXiv2020

Structure-based Sybil Detection in Social Networks via Local Rule-based Propagation

Sybil detection in social networks is a basic security research problem. Structure-based methods have been shown to be promising at detecting Sybils. Existing structure-based methods can be classified into Random Walk (RW)-based methods and Loop Belief Propagation (LBP)-based methods. RW-based methods cannot leverage labeled Sybils and labeled benign users simultaneously, which limits their detection accuracy, and/or they are not robust to noisy labels. LBP-based methods are not scalable and cannot guarantee convergence. In this work, we propose SybilSCAR, a novel structure-based method to detect Sybils in social networks. SybilSCAR is Scalable, Convergent, Accurate, and Robust to label noise. We first propose a framework to unify RW-based and LBP-based methods. Under our framework, these methods can be viewed as iteratively applying a (different) local rule to every user, which propagates label information among a social graph. Second, we design a new local rule, which SybilSCAR iteratively applies to every user to detect Sybils. We compare SybilSCAR with state-of-the-art RW-based and LBP-based methods theoretically and empirically. Theoretically, we show that, with proper parameter settings, SybilSCAR has a tighter asymptotical bound on the number of Sybils that are falsely accepted into a social network than existing structure-based methods. Empirically, we perform evaluation using both social networks with synthesized Sybils and a large-scale Twitter dataset (41.7M nodes and 1.2B edges) with real Sybils. Our results show that 1) SybilSCAR is substantially more accurate and more robust to label noise than state-of-the-art RW-based methods; 2) SybilSCAR is more accurate and one order of magnitude more scalable than state-of-the-art LBP-based methods.

cs.CR↗

arXiv2020

A novel centroid update approach for clustering-based superpixel methods and superpixel-based edge detection

Superpixel is widely used in image processing. And among the methods for superpixel generation, clustering-based methods have a high speed and a good performance at the same time. However, most clustering-based superpixel methods are sensitive to noise. To solve these problems, in this paper, we first analyze the features of noise. Then according to the statistical features of noise, we propose a novel centroid update approach to enhance the robustness of clustering-based superpixel methods. Besides, we propose a novel superpixel-based edge detection method. The experiments on BSD500 dataset show that our approach can significantly enhance the performance of clustering-based superpixel methods in noisy environment. Moreover, we also show that our proposed edge detection method outperforms other classical methods.

cs.CV↗