Searcharxiv⌕ Search

arXiv subjects

Yue Zhao

Publications and source records attributed to Yue Zhao.

At least 379 records · Page 21Linked to original sources
arXiv2021

A Large-scale Study on Unsupervised Outlier Model Selection: Do Internal Strategies Suffice?

Given an unsupervised outlier detection task, how should one select a detection algorithm as well as its hyperparameters (jointly called a model)? Unsupervised model selection is notoriously difficult, in the absence of hold-out validation data with ground-truth labels. Therefore, the problem is vastly understudied. In this work, we study the feasibility of employing internal model evaluation strategies for selecting a model for outlier detection. These so-called internal strategies solely rely on the input data (without labels) and the output (outlier scores) of the candidate models. We setup (and open-source) a large testbed with 39 detection tasks and 297 candidate models comprised of 8 detectors and various hyperparameter configurations. We evaluate 7 different strategies on their ability to discriminate between models w.r.t. detection performance, without using any labels. Our study reveals room for progress -- we find that none would be practically useful, as they select models only comparable to a state-of-the-art detector (with random configuration).

cs.LG↗
arXiv2021

HufuNet: Embedding the Left Piece as Watermark and Keeping the Right Piece for Ownership Verification in Deep Neural Networks

Due to the wide use of highly-valuable and large-scale deep neural networks (DNNs), it becomes crucial to protect the intellectual property of DNNs so that the ownership of disputed or stolen DNNs can be verified. Most existing solutions embed backdoors in DNN model training such that DNN ownership can be verified by triggering distinguishable model behaviors with a set of secret inputs. However, such solutions are vulnerable to model fine-tuning and pruning. They also suffer from fraudulent ownership claim as attackers can discover adversarial samples and use them as secret inputs to trigger distinguishable behaviors from stolen models. To address these problems, we propose a novel DNN watermarking solution, named HufuNet, for protecting the ownership of DNN models. We evaluate HufuNet rigorously on four benchmark datasets with five popular DNN models, including convolutional neural network (CNN) and recurrent neural network (RNN). The experiments demonstrate HufuNet is highly robust against model fine-tuning/pruning, kernels cutoff/supplement, functionality-equivalent attack, and fraudulent ownership claims, thus highly promising to protect large-scale DNN models in the real-world.

cs.CR↗
arXiv2021

Proximity Effect of Epitaxial Iron Phthalocyanine Molecules on High-Quality Graphene Devices

Depositing magnetic insulators on graphene has been a promising route to introduce magnetism via exchange proximity interaction in graphene for future spintronics applications. Molecule-based magnets may offer unique opportunities because of their synthesis versatility. Here, we investigated the magnetic proximity effect of epitaxial iron phthalocyanine (FePc) molecules on high-quality monolayer and bilayer graphene devices on hexagonal boron nitride substrate by probing the local and non-local transport. Although the FePc molecules introduce large hole doping effects combined with mobility degradation, the magnetic proximity gives rise to a canted antiferromagnetic state under a magnetic field in the monolayer graphene. On bilayer graphene and FePc heterostructure devices, the non-local transport reveals a pronounced Zeeman spin-Hall effect. Further analysis of the scattering mechanism in the bilayer shows a dominated long-range scattering. Our findings in graphene/organic magnetic insulator heterostructure provide a new insight for the use of molecule-based magnets in two-dimensional spintronic devices.

cond-mat.mes-hall↗
arXiv2021

Automating Outlier Detection via Meta-Learning

Given an unsupervised outlier detection (OD) task on a new dataset, how can we automatically select a good outlier detection method and its hyperparameter(s) (collectively called a model)? Thus far, model selection for OD has been a "black art"; as any model evaluation is infeasible due to the lack of (i) hold-out data with labels, and (ii) a universal objective function. In this work, we develop the first principled data-driven approach to model selection for OD, called MetaOD, based on meta-learning. MetaOD capitalizes on the past performances of a large body of detection models on existing outlier detection benchmark datasets, and carries over this prior experience to automatically select an effective model to be employed on a new dataset without using any labels. To capture task similarity, we introduce specialized meta-features that quantify outlying characteristics of a dataset. Through comprehensive experiments, we show the effectiveness of MetaOD in selecting a detection model that significantly outperforms the most popular outlier detectors (e.g., LOF and iForest) as well as various state-of-the-art unsupervised meta-learners while being extremely fast. To foster reproducibility and further research on this new problem, we open-source our entire meta-learning system, benchmark environment, and testbed datasets.

cs.LG↗
arXiv2021

A random batch Ewald method for particle systems with Coulomb interactions

We develop a random batch Ewald (RBE) method for molecular dynamics simulations of particle systems with long-range Coulomb interactions, which achieves an $O(N)$ complexity in each step of simulating the $N$-body systems. The RBE method is based on the Ewald splitting for the Coulomb kernel with a random "mini-batch" type technique introduced to speed up the summation of the Fourier series for the long-range part of the splitting. Importance sampling is employed to reduce the induced force variance by taking advantage of the fast decay property of the Fourier coefficients. The stochastic approximation is unbiased with controlled variance. Analysis for bounded force fields gives some theoretic support of the method. Simulations of two typical problems of charged systems are presented to illustrate the accuracy and efficiency of the RBE method in comparison to the results from the Debye-Hückel theory and the classical Ewald summation, demonstrating that the proposed method has the attractiveness of being easy to implement with the linear scaling and is promising for many practical applications.

physics.comp-ph↗
arXiv2021

Ultralight dark matter detection with mechanical quantum sensors

We consider the use of quantum-limited mechanical force sensors to detect ultralight (sub-meV) dark matter candidates which are weakly coupled to the standard model. We show that mechanical sensors with masses around or below the milligram scale, operating around the standard quantum limit, would enable novel searches for dark matter with natural frequencies around the kHz scale. This would complement existing strategies based on torsion balances, atom interferometers, and atomic clock systems.

hep-ph↗
arXiv2021

A 2 day orbital period for a redback millisecond pulsar candidate in the globular cluster NGC 6397

We report optical modulation of the companion to the X-ray source U18 in the globular cluster NGC 6397. U18, with combined evidence from radio and X-ray measurements, is a strong candidate as the second redback in this cluster, initially missed in pulsar searches. This object is a bright variable star with an anomalous red color and optical variability (\sim 0.2 mag in amplitude) with a periodicity \sim 1.96 days that can be interpreted as the orbital period. This value corresponds to the longest orbital period for known redback candidates and confirmed systems in Galactic globular clusters and one of the few with a period longer than 1 day.

astro-ph.HE↗
arXiv2021

SUOD: Accelerating Large-Scale Unsupervised Heterogeneous Outlier Detection

Outlier detection (OD) is a key machine learning (ML) task for identifying abnormal objects from general samples with numerous high-stake applications including fraud detection and intrusion detection. Due to the lack of ground truth labels, practitioners often have to build a large number of unsupervised, heterogeneous models (i.e., different algorithms with varying hyperparameters) for further combination and analysis, rather than relying on a single model. How to accelerate the training and scoring on new-coming samples by outlyingness (referred as prediction throughout the paper) with a large number of unsupervised, heterogeneous OD models? In this study, we propose a modular acceleration system, called SUOD, to address it. The proposed system focuses on three complementary acceleration aspects (data reduction for high-dimensional data, approximation for costly models, and taskload imbalance optimization for distributed environment), while maintaining performance accuracy. Extensive experiments on more than 20 benchmark datasets demonstrate SUOD's effectiveness in heterogeneous OD acceleration, along with a real-world deployment case on fraudulent claim analysis at IQVIA, a leading healthcare firm. We open-source SUOD for reproducibility and accessibility.

cs.LG↗
arXiv2021

Stability for an inverse source problem of the damped biharmonic plate equation

This paper is concerned with the stability of the inverse source problem for the damped biharmonic plate equation in three dimensions. The stability estimate consists of the Lipschitz type data discrepancy and the high frequency tail of the source function, where the latter decreases as the upper bound of the frequency increases. The stability also shows exponential dependence on the constant damping coefficient. The analysis employs Carleman estimates and time decay estimates for the damped plate wave equation to obtain an exact observability bound and depends on the study of the resonance-free region and an upper bound of the resolvent of the biharmonic operator with respect to the complex wavenumber.

math.AP↗
arXiv2021

Stability for an inverse source problem of the biharmonic operator

In this paper, we study for the first time the stability of the inverse source problem for the biharmonic operator with a compactly supported potential in $\mathbb R^3$. Firstly, to connect the boundary data with the unknown source, we shall consider an eigenvalue problem for the bi-Schr$\ddot{\rm o}$dinger operator $Δ^2 + V(x)$ on a ball which contains the support of the potential $V$. We prove a Weyl-type law for the upper bounds of spherical normal derivatives of both the eigenfunctions $ϕ$ and their Laplacian $Δϕ$ corresponding to the bi-Schr$\ddot{\rm o}$dinger operator. This type of upper bounds was proved by Hassell and Tao for the Schr$\ddot{\rm o}$dinger operator. Secondly, we investigate the meromorphic continuation of the resolvent of the bi-Schr$\ddot{\rm o}$dinger operator and prove the existence of a resonance-free region and an estimate of $L^2_{\rm comp} - L^2_{\rm loc}$ type for the resolvent. As an application, we prove a bound of the analytic continuation of the data from the given data to the higher frequency data. Finally, we derive the stability estimate which consists of the Lipschitz type data discrepancy and the high frequency tail of the source function, where the latter decreases as the upper bound of the frequency increases.

math.AP↗
arXiv2021

Inorganic component imaging of aggregate glue droplets on spider orb webs by TOF-SIMS

In this review, we discuss the use of time-of-flight secondary-ion mass spectrometry (TOF-SIMS) technology for analyzing the viscous glue (is called aggregate glue droplets) of spider orb webs and examine the results obtained. Element distribution images of the aggregate glue droplets were observed by TOF-SIMS. A uniform element distribution is seen for suspended pristine aggregate glue droplets, and a differential spreading of aggregate glue components is seen for attached aggregate glue droplets. We also observed TOF-SIMS images of water in aggregate glue droplets, where water was observed to be consistent with the distribution of oozing salt. We also found that the alkali metal in the aggregate glue droplets showed similar characteristics by feeding cesium carbonate to spiders.

physics.atm-clus↗
arXiv2021

Chandra and HST Studies of Six Millisecond Pulsars in the Globular Cluster M13

We analyse 55 ks of Chandra X-ray observations of the Galactic globular cluster M13. Using the latest radio timing positions of six known millisecond pulsars (MSPs) in M13 from Wang et al. (2020), we detect confident X-ray counterparts to five of the six MSPs at X-ray luminosities of $L_X$(0.3-8 keV)$\sim 3 \times 10^{30} - 10^{31}~{\rm erg~s^{-1}}$, including the newly discovered PSR J1641+3627F. There are limited X-ray counts at the position of PSR J1641+3627A, for which we obtain an upper limit $L_X<1.3 \times 10^{30}~{\rm erg~s^{-1}}$. We analyse X-ray spectra of all six MSPs, which are well-described by either a single blackbody or a single power-law model. We also incorporate optical/UV imaging observations from the Hubble Space Telescope (HST) and find optical counterparts to PSR J1641+3627D and J1641+3627F. Our colour-magnitude diagrams indicate the latter contains a white dwarf, consistent with the properties suggested by radio timing observations. The counterpart to J1641+3627D is only visible in the V band; however, we argue that the companion to J1641+3627D is also a white dwarf, since we see a blackbody-like X-ray spectrum, while MSPs with nondegenerate companions generally show non-thermal X-rays from shocks between the pulsar and companion winds. Our work increases the sample of known X-ray and optical counterparts of MSPs in globular clusters.

astro-ph.HE↗
arXiv2021

PyHealth: A Python Library for Health Predictive Models

Despite the explosion of interest in healthcare AI research, the reproducibility and benchmarking of those research works are often limited due to the lack of standard benchmark datasets and diverse evaluation metrics. To address this reproducibility challenge, we develop PyHealth, an open-source Python toolbox for developing various predictive models on healthcare data. PyHealth consists of data preprocessing module, predictive modeling module, and evaluation module. The target users of PyHealth are both computer science researchers and healthcare data scientists. With PyHealth, they can conduct complex machine learning pipelines on healthcare datasets with fewer than ten lines of code. The data preprocessing module enables the transformation of complex healthcare datasets such as longitudinal electronic health records, medical images, continuous signals (e.g., electrocardiogram), and clinical notes into machine learning friendly formats. The predictive modeling module provides more than 30 machine learning models, including established ensemble trees and deep neural network-based approaches, via a unified but extendable API designed for both researchers and practitioners. The evaluation module provides various evaluation strategies (e.g., cross-validation and train-validation-test split) and predictive model metrics. With robustness and scalability in mind, best practices such as unit testing, continuous integration, code coverage, and interactive examples are introduced in the library's development. PyHealth can be installed through the Python Package Index (PyPI) or https://github.com/yzhao062/PyHealth .

cs.LG↗
arXiv2021

RNN-Test: Towards Adversarial Testing for Recurrent Neural Network Systems

While massive efforts have been investigated in adversarial testing of convolutional neural networks (CNN), testing for recurrent neural networks (RNN) is still limited and leaves threats for vast sequential application domains. In this paper, we propose an adversarial testing framework RNN-Test for RNN systems, focusing on the main sequential domains, not only classification tasks. First, we design a novel search methodology customized for RNN models by maximizing the inconsistency of RNN states to produce adversarial inputs. Next, we introduce two state-based coverage metrics according to the distinctive structure of RNNs to explore more inference logics. Finally, RNN-Test solves the joint optimization problem to maximize state inconsistency and state coverage, and crafts adversarial inputs for various tasks of different kinds of inputs. For evaluations, we apply RNN-Test on three sequential models of common RNN structures. On the tested models, the RNN-Test approach is demonstrated to be competitive in generating adversarial inputs, outperforming FGSM-based and DLFuzz-based methods to reduce the model performance more sharply with 2.78% to 32.5% higher success (or generation) rate. RNN-Test could also achieve 52.65% to 66.45% higher adversary rate on MNIST-LSTM model than relevant work testRNN. Compared with the neuron coverage, the proposed state coverage metrics as guidance excel with 4.17% to 97.22% higher success (or generation) rate.

cs.CL↗
arXiv2020

TSGCNet: Discriminative Geometric Feature Learning with Two-Stream GraphConvolutional Network for 3D Dental Model Segmentation

The ability to segment teeth precisely from digitized 3D dental models is an essential task in computer-aided orthodontic surgical planning. To date, deep learning based methods have been popularly used to handle this task. State-of-the-art methods directly concatenate the raw attributes of 3D inputs, namely coordinates and normal vectors of mesh cells, to train a single-stream network for fully-automated tooth segmentation. This, however, has the drawback of ignoring the different geometric meanings provided by those raw attributes. This issue might possibly confuse the network in learning discriminative geometric features and result in many isolated false predictions on the dental model. Against this issue, we propose a two-stream graph convolutional network (TSGCNet) to learn multi-view geometric information from different geometric attributes. Our TSGCNet adopts two graph-learning streams, designed in an input-aware fashion, to extract more discriminative high-level geometric representations from coordinates and normal vectors, respectively. These feature representations learned from the designed two different streams are further fused to integrate the multi-view complementary information for the cell-wise dense prediction task. We evaluate our proposed TSGCNet on a real-patient dataset of dental models acquired by 3D intraoral scanners, and experimental results demonstrate that our method significantly outperforms state-of-the-art methods for 3D shape segmentation.

cs.CV↗
arXiv2020

Analytical solution for the surface states of antiferromagnetic topological insulator MnBi$_2$Te$_4$

Recently, the intrinsic magnetic topological insulator MnBi$_2$Te$_4$ has attracted great attention. It has an out-of-plane antiferromagnetic order, which is believed to open a sizable energy gap in the surface states. This gap, however, was not always observable in the latest angle-resolved photoemission spectroscopy (ARPES) experiments. To address this issue, we analytically derive an effective model for the two-dimensional (2D) surface states by starting from a three-dimensional (3D) Hamiltonian for bulk MnBi$_2$Te$_4$ and taking into account the spatial profile of the bulk magnetization. Our calculations suggest that the diminished surface gap may be caused by a much smaller and more localized intralayer ferromagnetic order. In addition, we calculate the spatial distribution and penetration depth of the surface states, which indicates that the surface states are mainly embedded in the first two septuple layers from the terminating surface. From our analytical results, the influence of the bulk parameters on the surface states can be found explicitly. Furthermore, we derive a $\bf{k}\cdot \bf{p}$ model for MnBi$_2$Te$_4$ thin films and show the oscillation of the Chern number between odd and even septuple layers. Our results will be helpful for the ongoing explorations of the MnBi$_x$Te$_y$ family.

cond-mat.mes-hall↗
arXiv2020

Modulating Electronic Structure of Monolayer Transition Metal Dichalcogenides by Substitutional Nb-Doping

Modulating electronic structure of monolayer transition metal dichalcogenides (TMDCs) is important for many applications and doping is an effective way towards this goal, yet is challenging to control. Here we report the in-situ substitutional doping of niobium (Nb) into TMDCs with tunable concentrations during chemical vapour deposition. Taking monolayer WS2 as an example, doping Nb into its lattice leads to bandgap changes in the range 1.98 eV to 1.65 eV. Noteworthy, electrical transport measurements and density functional theory calculations show that the 4d electron orbitals of the Nb dopants contribute to the density of states of Nb-doped WS2 around the Fermi level, resulting in an n to p-type conversion. Nb-doping also reduces the energy barrier of hydrogen absorption in WS2, leading to an improved electrocatalytic hydrogen evolution performance. These results highlight the effectiveness of controlled doping in modulating the electronic structure of TMDCs and their use in electronic related applications.

cond-mat.mtrl-sci↗