arXiv subjects

Xin Wang

Publications and source records attributed to Xin Wang.

At least 1,333 records · Page 74Linked to original sources

arXiv2021

Variational quantum Gibbs state preparation with a truncated Taylor series

The preparation of quantum Gibbs state is an essential part of quantum computation and has wide-ranging applications in various areas, including quantum simulation, quantum optimization, and quantum machine learning. In this paper, we propose variational hybrid quantum-classical algorithms for quantum Gibbs state preparation. We first utilize a truncated Taylor series to evaluate the free energy and choose the truncated free energy as the loss function. Our protocol then trains the parameterized quantum circuits to learn the desired quantum Gibbs state. Notably, this algorithm can be implemented on near-term quantum computers equipped with parameterized quantum circuits. By performing numerical experiments, we show that shallow parameterized circuits with only one additional qubit can be trained to prepare the Ising chain and spin chain Gibbs states with a fidelity higher than 95%. In particular, for the Ising chain model, we find that a simplified circuit ansatz with only one parameter and one additional qubit can be trained to realize a 99% fidelity in Gibbs state preparation at inverse temperatures larger than 2.

quant-ph↗

arXiv2021

Variational Quantum Algorithms for Trace Distance and Fidelity Estimation

Estimating the difference between quantum data is crucial in quantum computing. However, as typical characterizations of quantum data similarity, the trace distance and quantum fidelity are believed to be exponentially-hard to evaluate in general. In this work, we introduce hybrid quantum-classical algorithms for these two distance measures on near-term quantum devices where no assumption of input state is required. First, we introduce the Variational Trace Distance Estimation (VTDE) algorithm. We in particular provide the technique to extract the desired spectrum information of any Hermitian matrix by local measurement. A novel variational algorithm for trace distance estimation is then derived from this technique, with the assistance of a single ancillary qubit. Notably, VTDE could avoid the barren plateau issue with logarithmic depth circuits due to a local cost function. Second, we introduce the Variational Fidelity Estimation (VFE) algorithm. We combine Uhlmann's theorem and the freedom in purification to translate the estimation task into an optimization problem over a unitary on an ancillary system with fixed purified inputs. We then provide a purification subroutine to complete the translation. Both algorithms are verified by numerical simulations and experimental implementations, exhibiting high accuracy for randomly generated mixed states.

quant-ph↗

arXiv2021

The mass-metallicity relation at cosmic noon in overdense environments: first results from the MAMMOTH-Grism HST slitless spectroscopic survey

The MAMMOTH-Grism slitless spectroscopic survey is a Hubble Space Telescope (HST) cycle-28 medium program, which is obtaining 45 orbits of WFC3/IR grism spectroscopy in the density peak regions of three massive galaxy protoclusters at $z=2-3$ discovered using the MAMMOTH technique. We introduce this survey by presenting the first measurement of the mass-metallicity relation (MZR) at high redshift in overdense environments via grism spectroscopy. From the completed MAMMOTH-Grism observations in the field of the BOSS1244 protocluster at $z=2.24\pm0.02$, We secure a sample of 36 protocluster member galaxies at $z\sim2.24$, showing strong nebular emission lines ([O III], H$β$ and [O II]) in their G141 spectra. Using the multi-wavelength broad-band deep imaging from HST and ground-based telescopes, we measure their stellar masses in the range of $[10^{9},10^{10.4}]M_\odot$, instantaneous star formation rates (SFR) from 10 to 240$M_\odot yr^{-1}$, and global gas-phase metallicities [$\frac{1}{3}$,1] of solar. Compared with similarly selected field galaxy sample at the same redshift, our galaxies show on average increased SFRs by $\sim$0.06dex and $\sim$0.18dex at $\sim$10$^{10.1}M_\odot$ and $\sim$10$^{9.8}M_\odot$, respectively. Using the stacked spectra of our sample galaxies, we derive the MZR in the BOSS1244 protocluster core as $12+\log({\rm O/H})=(0.136\pm0.018)\times\log(M_\ast/M_\odot)+(7.082\pm0.175)$, showing significantly shallower slope than that in the field. This shallow MZR slope is likely caused by the combined effects of efficient recycling of feedback-driven winds and cold-mode gas accretion in protocluster environments. The former effect helps low-mass galaxies residing in overdensities retain their metal production, whereas the latter effect dilutes the metal content of high-mass galaxies, making them more metal poor than their coeval field counterparts.

astro-ph.GA↗

arXiv2021

E-string Quantum Curve

In this work we study the quantisation of the Seiberg-Witten curve for the E-string theory compactified on a two-torus. We find that the resulting operator expression belongs to the class of elliptic quantum curves. It can be rephrased as an eigenvalue equation with eigenvectors corresponding to co-dimension 2 defect operators and eigenvalues to co-dimension 4 Wilson surfaces wrapping the elliptic curve, respectively. Moreover, the operator we find is a generalised version of the van Diejen operator arising in the study of elliptic integrable systems. Although the microscopic representation of the co-dimension 4 defect only furnishes an $\mathrm{SO}(16)$ flavour symmetry in the UV, we find an enhancement in the IR to representations in terms of affine $E_8$ characters. Finally, using the Nekrasov-Shatashvili limit of the E-string BPS partition function, we give a path integral derivation of the quantum curve.

hep-th↗

arXiv2021

A Unified Game-Theoretic Interpretation of Adversarial Robustness

This paper provides a unified view to explain different adversarial attacks and defense methods, i.e. the view of multi-order interactions between input variables of DNNs. Based on the multi-order interaction, we discover that adversarial attacks mainly affect high-order interactions to fool the DNN. Furthermore, we find that the robustness of adversarially trained DNNs comes from category-specific low-order interactions. Our findings provide a potential method to unify adversarial perturbations and robustness, which can explain the existing defense methods in a principle way. Besides, our findings also make a revision of previous inaccurate understanding of the shape bias of adversarially learned features.

cs.LG↗

arXiv2021

A Unified Game-Theoretic Interpretation of Adversarial Robustness

This paper provides a unified view to explain different adversarial attacks and defense methods, \emph{i.e.} the view of multi-order interactions between input variables of DNNs. Based on the multi-order interaction, we discover that adversarial attacks mainly affect high-order interactions to fool the DNN. Furthermore, we find that the robustness of adversarially trained DNNs comes from category-specific low-order interactions. Our findings provide a potential method to unify adversarial perturbations and robustness, which can explain the existing defense methods in a principle way. Besides, our findings also make a revision of previous inaccurate understanding of the shape bias of adversarially learned features.

cs.LG↗

arXiv2021

Mitigating Quantum Errors via Truncated Neumann Series

Quantum gates and measurements on quantum hardware are inevitably subject to hardware imperfections that lead to quantum errors. Mitigating such unavoidable errors is crucial to explore the power of quantum hardware better. In this paper, we propose a unified framework that can mitigate quantum gate and measurement errors in computing quantum expectation values utilizing the truncated Neumann series. The essential idea is to cancel the effect of quantum error by approximating its inverse via linearly combining quantum errors of different orders produced by sequential applications of the quantum devices with carefully chosen coefficients. Remarkably, the estimation error decays exponentially in the truncated order, and the incurred error mitigation overhead is independent of the system size, as long as the noise resistance of the quantum device is moderate. We numerically test this framework for different quantum errors and find that the computation accuracy is substantially improved. Our framework possesses several vital advantages: it mitigates quantum gate and measurement errors in a unified manner, it neither assumes any error structure nor requires the tomography procedure to completely characterize the quantum errors, and most importantly, it is scalable. These advantages empower our quantum error mitigation framework to be efficient and practical and extend the ability of near-term quantum devices to deliver quantum applications.

quant-ph↗

arXiv2021

CAP: Co-Adversarial Perturbation on Weights and Features for Improving Generalization of Graph Neural Networks

Despite the recent advances of graph neural networks (GNNs) in modeling graph data, the training of GNNs on large datasets is notoriously hard due to the overfitting. Adversarial training, which augments data with the worst-case adversarial examples, has been widely demonstrated to improve model's robustness against adversarial attacks and generalization ability. However, while the previous adversarial training generally focuses on protecting GNNs from spiteful attacks, it remains unclear how the adversarial training could improve the generalization abilities of GNNs in the graph analytics problem. In this paper, we investigate GNNs from the lens of weight and feature loss landscapes, i.e., the loss changes with respect to model weights and node features, respectively. We draw the conclusion that GNNs are prone to falling into sharp local minima in these two loss landscapes, where GNNs possess poor generalization performances. To tackle this problem, we construct the co-adversarial perturbation (CAP) optimization problem in terms of weights and features, and design the alternating adversarial perturbation algorithm to flatten the weight and feature loss landscapes alternately. Furthermore, we divide the training process into two stages: one conducting the standard cross-entropy minimization to ensure the quick convergence of GNN models, the other applying our alternating adversarial training to avoid falling into locally sharp minima. The extensive experiments demonstrate our CAP can generally improve the generalization performance of GNNs on a variety of benchmark graph datasets.

cs.LG↗

arXiv2021

RF-Based Human Activity Recognition Using Signal Adapted Convolutional Neural Network

Human Activity Recognition (HAR) plays a critical role in a wide range of real-world applications, and it is traditionally achieved via wearable sensing. Recently, to avoid the burden and discomfort caused by wearable devices, device-free approaches exploiting RF signals arise as a promising alternative for HAR. Most of the latest device-free approaches require training a large deep neural network model in either time or frequency domain, entailing extensive storage to contain the model and intensive computations to infer activities. Consequently, even with some major advances on device-free HAR, current device-free approaches are still far from practical in real-world scenarios where the computation and storage resources possessed by, for example, edge devices, are limited. Therefore, we introduce HAR-SAnet which is a novel RF-based HAR framework. It adopts an original signal adapted convolutional neural network architecture: instead of feeding the handcraft features of RF signals into a classifier, HAR-SAnet fuses them adaptively from both time and frequency domains to design an end-to-end neural network model. We apply point-wise grouped convolution and depth-wise separable convolutions to confine the model scale and to speed up the inference execution time. The experiment results show that the recognition accuracy of HAR-SAnet outperforms state-of-the-art algorithms and systems.

cs.NI↗

arXiv2021

Elliptic Quantum Curves of 6d SO(N) theories

We discuss supersymmetric defects in 6d $\mathcal{N}=(1,0)$ SCFTs with $\mathrm{SO}(N_c)$ gauge group and $N_c-8$ fundamental flavors. The codimension 2 and 4 defects are engineered by coupling the 6d gauge fields to charged free fields in four and two dimensions, respectively. We find that the partition function in the presence of the codimension 2 defect on $\mathbb{R}^4\times \mathbb{T}^2$ in the Nekrasov-Shatashvili limit satisfies an elliptic difference equation which quantizes the Seiberg-Witten curve of the 6d theory. The expectation value of the codimension 4 defect appearing in the difference equation is an even (under reflection) degree $N_c$ section over the elliptic curve when $N_c$ is even, and an odd section when $N_c$ is odd. We also find that RG-flows of the defects and the associated difference equations in the 6d $\mathrm{SO}(2N+1)$ gauge theories triggered by Higgs VEVs of KK-momentum states provide quantum Seiberg-Witten curves for $\mathbb{Z}_2$ twisted compactifications of the 6d $\mathrm{SO}(2N)$ gauge theories.

hep-th↗

arXiv2021

Pursuing the fundamental limits for quantum communication

The quantum capacity of a noisy quantum channel determines the maximal rate at which we can code reliably over asymptotically many uses of the channel, and it characterizes the channel's ultimate ability to transmit quantum information coherently. In this paper, we derive single-letter upper bounds on the quantum and private capacities of quantum channels. The quantum capacity of a quantum channel is always no larger than the quantum capacity of its extended channels, since the extensions of the channel can be considered as assistance from the environment. By optimizing the parametrized extended channels with specific structures such as the flag structure, we obtain new upper bounds on the quantum capacity of the original quantum channel. Furthermore, we extend our approach to estimating the fundamental limits of private communication and one-way entanglement distillation. As notable applications, we establish improved upper bounds to the quantum and private capacities for fundamental quantum channels of interest in quantum information, some of which are also the sources of noise in superconducting quantum computing. In particular, our upper bounds on the quantum capacities of the depolarizing channel and the generalized amplitude damping channel are strictly better than previously best-known bounds for certain regimes.

quant-ph↗

arXiv2021

Over-the-Air Federated Multi-Task Learning

In this letter, we introduce over-the-air computation into the communication design of federated multi-task learning (FMTL), and propose an over-the-air federated multi-task learning (OA-FMTL) framework, where multiple learning tasks deployed on edge devices share a non-orthogonal fading channel under the coordination of an edge server (ES). Specifically, the model updates for all the tasks are transmitted and superimposed concurrently over a non-orthogonal uplink fading channel, and the model aggregations of all the tasks are reconstructed at the ES through a modified version of the turbo compressed sensing algorithm (Turbo-CS) that overcomes inter-task interference. Both convergence analysis and numerical results show that the OA-FMTL framework can significantly improve the system efficiency in terms of reducing the number of channel uses without causing substantial learning performance degradation.

cs.LG↗

arXiv2021

The Origin of Radio Emission in Black Hole X-ray Binaries

We studied the relation of accretion-jet power and disk luminosity, especially the jet efficiencies and disk radiative efficiencies for different accretion disks as well as black hole (BH) spin, in order to explore the origin of radio emission in black hole X-ray binaries (BHXBs). We found that jet efficiency increases more rapidly (efficient) than the nearly constant disk radiative efficiency for thin disk component in high accretion regime, which could account for the steep track ($μ>1$) in the observed radio and X-ray luminosity relations ($L_{\rm R}\propto L_{\rm X}^μ$), but the thin disk component may not be able to explain the standard track ($μ\approx 0.6$) in the BHXBs. For hot accretion flows (HAF), the resulting jet efficiency changes along with the large range of accretions from quiescent state to nearly Eddington state, which could account for the standard track in the BHXBs. The BH spin-jet is discussed for the magnetic arrested disk (MAD) state; in this state, the spin-jet power might contribute to a linear correlation between jet power and mass accretion rate for a given source. More accurate observations are required to test the results.

astro-ph.HE↗

arXiv2021

Evaluation of Various Open-Set Medical Imaging Tasks with Deep Neural Networks

The current generation of deep neural networks has achieved close-to-human results on "closed-set" image recognition; that is, the classes being evaluated overlap with the training classes. Many recent methods attempt to address the importance of the unknown, which are termed "open-set" recognition algorithms, try to reject unknown classes as well as maintain high recognition accuracy on known classes. However, it is still unclear how different general domain-trained open-set methods from ImageNet would perform on a different but more specific domain, such as the medical domain. Without principled and formal evaluations to measure the effectiveness of those general open-set methods, artificial intelligence (AI)-based medical diagnostics would experience ineffective adoption and increased risks of bad decision making. In this paper, we conduct rigorous evaluations amongst state-of-the-art open-set methods, exploring different open-set scenarios from "similar-domain" to "different-domain" scenarios and comparing them on various general and medical domain datasets. We summarise the results and core ideas and explain how the models react to various degrees of openness and different distributions of open classes. We show the main difference between general domain-trained and medical domain-trained open-set models with our quantitative and qualitative analysis of the results. We also identify aspects of model robustness in real clinical workflow usage according to confidence calibration and the inference efficiency.

eess.IV↗

arXiv2021

Efficiently Solve the Max-cut Problem via a Quantum Qubit Rotation Algorithm

Optimizing parameterized quantum circuits promises efficient use of near-term quantum computers to achieve the potential quantum advantage. However, there is a notorious tradeoff between the expressibility and trainability of the parameter ansatz. We find that in combinatorial optimization problems, since the solutions are described by bit strings, one can trade the expressiveness of the ansatz for high trainability. To be specific, by focusing on the max-cut problem we introduce a simple yet efficient algorithm named Quantum Qubit Rotation Algorithm (QQRA). The quantum circuits are comprised with single-qubit rotation gates implementing on each qubit. The rotation angles of the gates can be trained free of barren plateaus. Thus, the approximate solution of the max-cut problem can be obtained with probability close to 1. To illustrate the effectiveness of QQRA, we compare it with the well known quantum approximate optimization algorithm and the classical Goemans-Williamson algorithm.

quant-ph↗

arXiv2021

Machine learning of XY model on a spherical Fibonacci lattice

We study the XY model on a spherical surface inspired by recently realized spherically confined atomic gases. Instead of a traditional latitude-longitude lattice, we introduce a much more homogeneous spherical lattice, the Fibonacci lattice, and use classical Monte Carlo simulations to determine spin configurations. The results clearly show that topological defects, in the form of vortices, must exist in the stable configuration on a sphere but vanish in a plane due to a mathematical theorem. Using these spin configurations as training samples, we propose a graph-convolutional-network based method to recognize different phases, and successfully predict the phase transition temperature. We also apply the density-based spatial clustering of applications with noise, a powerful machine learning algorithm, to monitor the merging path of two vortices with different topological charges on the sphere during Monte Carlo simulations. Our results provide reliable predictions for future space-based experiments on ultracold atomic gases confined on spherical lattice in the microgravity environment.

cond-mat.quant-gas↗

arXiv2021

Design Strength-Ductility Synergy of Metastable High-Entropy Alloys by Tailoring Unstable Fault Energies

Metastable alloys with transformation/twinning-induced plasticity (TRIP/TWIP) can overcome the strength-ductility trade-off in structural materials. Originated from the development of traditional alloys, the intrinsic stacking fault energy (ISFE) has been relied to tailor TRIP/TWIP in high-entropy alloys (HEA), but with limited quantitative success. Herein, we demonstrate a new strategy for designing metastable HEAs and validate its effectiveness by discovering seven new alloys with experimentally observed metastability for TRIP/TWIP. We propose unstable fault energies as the more effective design metric and attribute the deformation mechanism of metastable face-centered cubic alloys to UMFE (unstable martensite fault energy)/UTFE (unstable twin fault energy) rather than ISFE. Among the studied HEAs and steels, the traditional ISFE criterion fails in more than half of the cases, while the UMFE/UTFE criterion accurately predicts the deformation mechanisms in all cases. The UMFE/UTFE criterion provides a new paradigm for developing metastable alloys with TRIP/TWIP for enhanced strength-ductility synergy.

cond-mat.mtrl-sci↗

arXiv2021

Improved Lower Bounds for Strongly Separable Matrices and Related Combinatorial Structures

In nonadaptive group testing, the main research objective is to design an efficient algorithm to identify a set of up to $t$ positive elements among $n$ samples with as few tests as possible. Disjunct matrices and separable matrices are two classical combinatorial structures while one provides a more efficient decoding algorithm and the other needs fewer tests, i.e., larger rate. Recently, a notion of strongly separable matrix has been introduced, which has the same identifying ability as a disjunct matrix, but has larger rate. In this paper, we use a modified probabilistic method to improve the lower bounds for the rate of strongly separable matrices. Using this method, we also improve the lower bounds for some well-known combinatorial structures, including locally thin set families and cancellative set families.

math.CO↗