arXiv subjects

Peng Li

Publications and source records attributed to Peng Li.

At least 469 records · Page 26Linked to original sources

arXiv2019

Probing the Origin of Extreme Magnetoresistance in Pr/Sm Mono-Antimonides/Bismuthides

Combining angle-resolved photoemission spectroscopy and magneto-transport measurements, we systematically investigated the possible origin of the extreme magnetoresistance in Pr/Sm mono-antimonides/bismuthides (PrSb, SmSb, PrBi, SmBi). Our photoemission measurements reveal that the bulk band inversion and surface states are absent (present) in Pr/Sm antimonides (bismuthides), implying that topological surface states are unlikely to play an important role for the observed extreme magnetoresistance. We found that the electron-hole compensation is well satisfied in all these compounds and the bulk band structure exhibits no obvious temperature dependence from 10 K up to 150 K. Simultaneous fittings of the magnetoresistance and Hall coefficient reveal that the carrier mobility is dramatically enhanced at low temperature, which naturally explains the suppression of extreme magnetoresistance at high temperatures. Our results therefore show that the extreme magnetoresistance in these compounds can be well accounted for by the two-band model with good electron-hole compensation. Finally, we found that both PrSb and SmSb exhibit highly linear bulk bands near the X point and lie close to the transition point between a topologically trivial and nontrivial phase, which might be relevant for the observed anomalous quantum oscillations.

cond-mat.mes-hall↗

arXiv2019

Hybrid Macro/Micro Level Backpropagation for Training Deep Spiking Neural Networks

Spiking neural networks (SNNs) are positioned to enable spatio-temporal information processing and ultra-low power event-driven neuromorphic hardware. However, SNNs are yet to reach the same performances of conventional deep artificial neural networks (ANNs), a long-standing challenge due to complex dynamics and non-differentiable spike events encountered in training. The existing SNN error backpropagation (BP) methods are limited in terms of scalability, lack of proper handling of spiking discontinuities, and/or mismatch between the rate-coded loss function and computed gradient. We present a hybrid macro/micro level backpropagation (HM2-BP) algorithm for training multi-layer SNNs. The temporal effects are precisely captured by the proposed spike-train level post-synaptic potential (S-PSP) at the microscopic level. The rate-coded errors are defined at the macroscopic level, computed and back-propagated across both macroscopic and microscopic levels. Different from existing BP methods, HM2-BP directly computes the gradient of the rate-coded loss function w.r.t tunable parameters. We evaluate the proposed HM2-BP algorithm by training deep fully connected and convolutional SNNs based on the static MNIST [14] and dynamic neuromorphic N-MNIST [26]. HM2-BP achieves an accuracy level of 99.49% and 98.88% for MNIST and N-MNIST, respectively, outperforming the best reported performances obtained from the existing SNN BP algorithms. Furthermore, the HM2-BP produces the highest accuracies based on SNNs for the EMNIST [3] dataset, and leads to high recognition accuracy for the 16-speaker spoken English letters of TI46 Corpus [16], a challenging patio-temporal speech recognition benchmark for which no prior success based on SNNs was reported. It also achieves competitive performances surpassing those of conventional deep learning models when dealing with asynchronous spiking streams.

cs.NE↗

arXiv2018

Ring frustration and factorizable correlation functions of critical spin rings

Basing on the exactly solvable prototypical model, the critical transverse Ising ring with or without ring frustration, we establish the concept of nonlocality in a many-body system in the thermodynamic limit by defining the nonlocal factors embedded in its factorizable correlation functions. In the context of nonlocality, the valuable traditional finite-size scaling analysis is reappraised. The factorizable correlation functions of the isotropic $XY$ and the spin-1/2 Heisenberg models are also demonstrated with the emphasis on the effect of ring frustration.

cond-mat.stat-mech↗

arXiv2018

Efficient and Robust Recovery of Signal and Image in Impulsive Noise via $\ell_1-α\ell_2$ Minimization

In this paper, we consider the efficient and robust reconstruction of signals and images via $\ell_{1}-α\ell_{2}~(0<α\leq 1)$ minimization in impulsive noise case. To achieve this goal, we introduce two new models: the $\ell_1-α\ell_2$ minimization with $\ell_1$ constraint, which is called $\ell_1-α\ell_2$-LAD, the $\ell_1-α\ell_2$ minimization with Dantzig selector constraint, which is called $\ell_1-α\ell_2$-DS. We first show that sparse signals or nearly sparse signals can be exactly or stably recovered via $\ell_{1}-α\ell_{2}$ minimization under some conditions based on the restricted $1$-isometry property ($\ell_1$-RIP). Second, for $\ell_1-α\ell_2$-LAD model, we introduce unconstrained $\ell_1-α\ell_2$ minimization model denoting $\ell_1-α\ell_2$-PLAD and propose $\ell_1-α\ell_2$LA algorithm to solve the $\ell_1-α\ell_2$-PLAD. Last, numerical experiments %on success rates of sparse signal recovery demonstrate that when the sensing matrix is ill-conditioned (i.e., the coherence of the matrix is larger than 0.99), the $\ell_1-α\ell_2$LA method is better than the existing convex and non-convex compressed sensing solvers for the recovery of sparse signals. And for the magnetic resonance imaging (MRI) reconstruction with impulsive noise, we show that the $\ell_1-α\ell_2$LA method has better performance than state-of-the-art methods via numerical experiments.

math.OC↗

arXiv2018

Energy-resolved Photoconductivity Mapping in a Monolayer-bilayer WSe2 Lateral Heterostructure

Vertical and lateral heterostructures of van der Waals materials provide tremendous flexibility for band structure engineering. Since electronic bands are sensitively affected by defects, strain, and interlayer coupling, the edge and heterojunction of these two-dimensional (2D) systems may exhibit novel physical properties, which can be fully revealed only by spatially resolved probes. Here, we report the spatial mapping of photoconductivity in a monolayer-bilayer WSe2 lateral heterostructure under multiple excitation lasers. As the photon energy increases, the light-induced conductivity detected by microwave impedance microscopy first appears along the hetero-interface and bilayer edge, then along the monolayer edge, inside the bilayer area, and finally in the interior of the monolayer region. The sequential emergence of mobile carriers in different sections of the sample is consistent with the theoretical calculation of local energy gaps. Quantitative analysis of the microscopy and transport data also reveals the linear dependence of photoconductivity on the laser intensity and the influence of interlayer coupling on carrier recombination. Combining theoretical modeling, atomic scale imaging, mesoscale impedance microscopy, and device-level characterization, our work suggests an exciting perspective to control the intrinsic band-gap variation in 2D heterostructures down to the few-nanometer regime.

cond-mat.mtrl-sci↗

arXiv2018

Efficient Two-Step Adversarial Defense for Deep Neural Networks

In recent years, deep neural networks have demonstrated outstanding performance in many machine learning tasks. However, researchers have discovered that these state-of-the-art models are vulnerable to adversarial examples: legitimate examples added by small perturbations which are unnoticeable to human eyes. Adversarial training, which augments the training data with adversarial examples during the training process, is a well known defense to improve the robustness of the model against adversarial attacks. However, this robustness is only effective to the same attack method used for adversarial training. Madry et al.(2017) suggest that effectiveness of iterative multi-step adversarial attacks and particularly that projected gradient descent (PGD) may be considered the universal first order adversary and applying the adversarial training with PGD implies resistance against many other first order attacks. However, the computational cost of the adversarial training with PGD and other multi-step adversarial examples is much higher than that of the adversarial training with other simpler attack techniques. In this paper, we show how strong adversarial examples can be generated only at a cost similar to that of two runs of the fast gradient sign method (FGSM), allowing defense against adversarial attacks with a robustness level comparable to that of the adversarial training with multi-step adversarial examples. We empirically demonstrate the effectiveness of the proposed two-step defense approach against different attack methods and its improvements over existing defense strategies.

cs.LG↗

arXiv2018

Optimized Gated Deep Learning Architectures for Sensor Fusion

Sensor fusion is a key technology that integrates various sensory inputs to allow for robust decision making in many applications such as autonomous driving and robot control. Deep neural networks have been adopted for sensor fusion in a body of recent studies. Among these, the so-called netgated architecture was proposed, which has demonstrated improved performances over the conventional convolutional neural networks (CNN). In this paper, we address several limitations of the baseline negated architecture by proposing two further optimized architectures: a coarser-grained gated architecture employing (feature) group-level fusion weights and a two-stage gated architectures leveraging both the group-level and feature level fusion weights. Using driving mode prediction and human activity recognition datasets, we demonstrate the significant performance improvements brought by the proposed gated architectures and also their robustness in the presence of sensor noise and failures.

cs.LG↗

arXiv2018

Giant Planar Hall Effect in the Dirac Semimetal ZrTe5

Recently, giant planar Hall effect originating from chiral anomaly has been predicted in nonmagnetic Dirac/Weyl semimetals. ZrTe5 is considered to be an intriguing Dirac semimetal at the boundary of weak topological insulators and strong topological insulators, though this claim still remains controversial. Here, we report the observation in ZrTe5 of the giant planar Hall resistivity that shows two different magnetic-field dependences as predicted by theory and a maximum at the Lifshitz transition temperature. We found that the giant planar Hall resistivity fades out with decreasing the thickness of ZrTe5 nanoplates, which may be ascribed to the vanishing of the 3D nature of the samples. In addition, we have observed a nontrivial Berry phase, chiral-anomaly-induced negative longitudinal magnetoresistance, and a giant in-plane anisotropic magnetoresistance in these ZrTe5 nanoplates. All the experimental observations demonstrated coherently that ZrTe5 is a Dirac semimetal.

cond-mat.mes-hall↗

arXiv2018

CAPRL: Signal Recovery from Compressive Affine Phase Retrieval via Lifting

In this paper, we consider compressive/sparse affine phase retrieval proposed in [B. Gao B, Q. Sun, Y. Wang and Z. Xu, Adv. in Appl. Math., 93(2018), 121-141]. By the lift technique, and heuristic nuclear norm for convex relaxation of rank and $\ell$ one norm convex relaxation of sparsity, we establish convex models , which are called compressive affine phase retrieval via lifting (CAPRL). In order to compute these models, we develop inertial proximal ADMM for multiple separated operators and also give out its convergence analysis. Our numerical experiments via proposed algorithm show that sparse signal can be exactly and stably recovered via CAPRL. We also list some other applications of our proposed algorithm.

math.OC↗

arXiv2018

Image Captioning based on Deep Reinforcement Learning

Recently it has shown that the policy-gradient methods for reinforcement learning have been utilized to train deep end-to-end systems on natural language processing tasks. What's more, with the complexity of understanding image content and diverse ways of describing image content in natural language, image captioning has been a challenging problem to deal with. To the best of our knowledge, most state-of-the-art methods follow a pattern of sequential model, such as recurrent neural networks (RNN). However, in this paper, we propose a novel architecture for image captioning with deep reinforcement learning to optimize image captioning tasks. We utilize two networks called "policy network" and "value network" to collaboratively generate the captions of images. The experiments are conducted on Microsoft COCO dataset, and the experimental results have verified the effectiveness of the proposed method.

cs.CV↗

arXiv2018

Multilinear Operators on Weighted Amalgam-Type Spaces

In this paper, we prove that if a multilinear operator $\mathcal{T}$ and its multilinear commutator $\mathcal{T}_{Σ\vec{b}}$ and iterated commutator $\mathcal{T}_{Π\vec{b}}$ for $\vec{b}\in(\mathbb{R}^n)^m$ are bounded on product weighted Lebesgue space, then $\mathcal{T}$, $\mathcal{T}_{Σ\vec{b}}$ and $\mathcal{T}_{Π\vec{b}}$ are also bounded on product weighted Amalgam space. As its applications, we show that multilinear Littlewood-Paley functions and multilinear Marcinkiewicz integral functions with kernels of convolution type and non-convolution type, and their multilinear commutators and iterated commutators are bounded on product weighted Amalgam space. We also consider multilinear fractional type integral operators and their commutators' behaviors on weighted amalgam space. In order to deal with the endpoint case, we introduce the amalgam-Campanato space and show that fractional integral integral operator are bounded operators from product amalgam space to amalgam-Campanato space. What should point out is that even if in the linear case, our results for fractional integral operator are also new.

math.FA↗

arXiv2018

The A-Cycle Problem for Transverse Ising Ring

Traditionally, the transverse Ising model is mapped to the fermionic c-cycle problem, which neglects the boundary effect due to thermodynamic limit. If persisting on a perfect periodic boundary condition, we can get a so-called a-cycle problem that has not been treated seriously so far (Lieb et al., 1961 \textit{Ann. of Phys.} \textbf{16} 407). In this work, we show a little surprising but exact result in this respect. We find the odevity of the number of lattice sites, $N$, in the a-cycle problem plays an unexpected role even in the thermodynamic limit, $N\rightarrow\infty$, due to the boundary constraint. We pay a special attention to the system with $N(\in Odd)\rightarrow\infty$, which is in contrast to the one with $N(\in Even)\rightarrow\infty$, because the former suffers a ring frustration. As a new effect, we find the ring frustration induces a low-energy gapless spectrum above the ground state. By proving a theorem for a new type of Toeplitz determinant, we demonstrate that the ground state in the gapless region exhibits a peculiar longitudinal spin-spin correlation. The entangled nature of the ground state is also disclosed by the evaluation of its entanglement entropy. At low temperatures, new behavior of specific heat is predicted. We also propose an experimental protocol for observing the new phenomenon due to the ring frustration.

cond-mat.quant-gas↗

arXiv2018

Signal Recovery under Mutual Incoherence Property and Oracle Inequalities

This paper considers signal recovery through an unconstrained minimization in the framework of mutual incoherence property. A sufficient condition is provided to guarantee the stable recovery in the noisy case. And we give a lower bound for the $\ell_2$ norm of difference of reconstructed signals and the original signal, in the sense of expectation and probability. Furthermore, oracle inequalities of both sparse signals and non-sparse signals are derived under the mutual incoherence condition in the case of Gaussian noises. Finally, we investigate the relationship between mutual incoherence property and robust null space property and find that robust null space property can be deduced from the mutual incoherence property.

cs.IT↗

arXiv2018

Matrix Recovery from Rank-One Projection Measurements via Nonconvex Minimization

In this paper, we consider the matrix recovery from rank-one projection measurements proposed in [Cai and Zhang, Ann. Statist., 43(2015), 102-138], via nonconvex minimization. We establish a sufficient identifiability condition, which can guarantee the exact recovery of low-rank matrix via Schatten-$p$ minimization $\min_{X}\|X\|_{S_p}^p$ for $0<p<1$ under affine constraint, and stable recovery of low-rank matrix under $\ell_q$ constraint and Dantzig selector constraint. Our condition is also sufficient to guarantee low-rank matrix recovery via least $q$ minimization $\min_{X}\|\mathcal{A}(X)-b\|_{q}^q$ for $0<q\leq1$. And we also extend our result to Gaussian design distribution, and show that any matrix can be stably recovered for rank-one projection from Gaussian distributions via least $1$ minimization with high probability.

cs.IT↗

arXiv2018

Truncated Sparse Approximation Property and Truncated $q$-Norm Minimization

This paper considers approximately sparse signal and low-rank matrix's recovery via truncated norm minimization $\min_{x}\|x_T\|_q$ and $\min_{X}\|X_T\|_{S_q}$ from noisy measurements. We first introduce truncated sparse approximation property, a more general robust null space property, and establish the stable recovery of signals and matrices under the truncated sparse approximation property. We also explore the relationship between the restricted isometry property and truncated sparse approximation property. And we also prove that if a measurement matrix $A$ or linear map $\mathcal{A}$ satisfies truncated sparse approximation property of order $k$, then the first inequality in restricted isometry property of order $k$ and of order $2k$ can hold for certain different constants $δ_{k}$ and $δ_{2k}$, respectively. Last, we show that if $δ_{t(k+|T^c|)}<\sqrt{(t-1)/t}$ for some $t\geq 4/3$, then measurement matrix $A$ and linear map $\mathcal{A}$ satisfy truncated sparse approximation property of order $k$. Which should point out is that when $T^c=\emptyset$, our conclusion implies that sparse approximation property of order $k$ is weaker than restricted isometry property of order $tk$.

cs.IT↗

arXiv2018

Signal Recovery under Cumulative Coherence

This paper considers signal recovery in the framework of cumulative coherence. First, we show that the Lasso estimator and the Dantzig selector exhibit similar behavior under the cumulative coherence. Then we estimate the approximation equivalence between the Lasso and the Dantzig selector by calculating prediction loss difference under the condition of cumulative coherence. And we also prove that the cumulative coherence implies the restricted eigenvalue condition. Last, we illustrate the advantages of cumulative coherence condition for three class matrices, in terms of the recovery performance of sparse signals via extensive numerical experiments.

cs.IT↗

arXiv2018

Joint Transmitter and Receiver Design for Pattern Division Multiple Access

In this paper, a joint transmitter and receiver design for pattern division multiple access (PDMA) is proposed. At the transmitter, pattern mapping utilizes power allocation to improve the overall sum rate, and beam allocation to enhance the access connectivity. At the receiver, hybrid detection utilizes a spatial filter to suppress the inter-beam interference caused by beam domain multiplexing, and successive interference cancellation to remove the intra-beam interference caused by power domain multiplexing. Furthermore, we propose a PDMA joint design approach to optimize pattern mapping based on both the power domain and beam domain. The optimization of power allocation is achieved by maximizing the overall sum rate, and the corresponding optimization problem is shown to be convex theoretically. The optimization of beam allocation is achieved by minimizing the maximum of the inner product of any two beam allocation vectors, and an effective dimension reduction method is proposed through the analysis of pattern structure and proper mathematical manipulations. Simulation results show that the proposed PDMA approach outperforms the orthogonal multiple access and power-domain non-orthogonal multiple access approaches even without any optimization of pattern mapping, and that the optimization of beam allocation yields a significant performance improvement than the optimization of power allocation.

cs.IT↗

arXiv2017

In situ high resolution real-time quantum efficiency imaging for photocathodes

Aspects of the preparation process and performance degradation are two major problems of photocathodes. The lack of a means for dynamic quantum efficiency measurements results in the inability to observe the inhomogeneity of the cathode surface by fine structural analysis and in real time.Here we present a simple and scalable technique for in situ real-time quantum efficiency diagnosis. An incoherent light source provides uniform illumination on the cathode surface, and solenoid magnets are used as lens for focusing and imaging the emitted electron beam on a downstream scintillator screen, which converts the quantum efficiency information into fluorescence intensity distribution. The microscopic discontinuity and the dynamic changes of the quantum efficiency of a gallium arsenide photocathode are observed at a resolution of a few microns. An unexpected uneven decrease of the quantum efficiency is also recorded. The work demonstrates a new observation method for photoemission materials research.

physics.app-ph↗