arXiv subjects

Zhendong Zhang

Publications and source records attributed to Zhendong Zhang.

26 records · Page 2Linked to original sources

arXiv2020

Jet Sub-structure in Fireworks Emission from Non-uniform and Rotating Bose-Einstein Condensates

We show that jet emission from a Bose condensate with periodically driven interactions, a.k.a. "Bose fireworks", contains essential information on the condensate wavefunction, which is difficult to obtain using standard detection methods. We illustrate the underlying physics with two examples. When condensates acquire phase patterns from external potentials or from vortices, the jets display novel sub-structure, such as oscillations or spirals, in their correlations. Through a comparison of theory, numerical simulations and experiments, we show how one can quantitatively extract the phase and the helicity of a condensate from the emission pattern. Our work demonstrating the strong link between jet emission and the underlying quantum system, bears on the recent emphasis on jet sub-structure in particle physics.

cond-mat.quant-gas↗

arXiv2020

Voltage-Induced Inertial Domain Wall Motion in an Antiferromagnetic Nanowire

Racetrack memory based on magnetic domain walls (DWs) motion exhibits advantages of small volume and high reading speed. When compared to current-induced DW motion, voltage-induced DW motion exhibits lower dissipation. On the other hand, the DW in an antiferromagnet (AFM) moves at a high velocity with weak stray field. In this work, the AFM DW motion induced by a gradient of magnetic anisotropy energy under a voltage pulse has been investigated in theory. The dynamics equation for the DW motion was derived. The solution indicates that the DW velocity is higher than 100 m/s, and because of inertia, the DW is able to keep moving at a speed of around 100 m/s for several nano seconds after turning off the voltage in a period of pulse. The mechanism for this DW inertia is explained based on the Lagrangian route. On the other hand, a spin wave is emitted while the DW is moving, yet the DW is still able to move at an ever increasing velocity with enlarging DW width. This indicates energy loss from emission of spin wave is less than the energy gain from the effective field of the gradient of anisotropy energy.

physics.app-ph↗

arXiv2019

GBDT-MO: Gradient Boosted Decision Trees for Multiple Outputs

Gradient boosted decision trees (GBDTs) are widely used in machine learning, and the output of current GBDT implementations is a single variable. When there are multiple outputs, GBDT constructs multiple trees corresponding to the output variables. The correlations between variables are ignored by such a strategy causing redundancy of the learned tree structures. In this paper, we propose a general method to learn GBDT for multiple outputs, called GBDT-MO. Each leaf of GBDT-MO constructs predictions of all variables or a subset of automatically selected variables. This is achieved by considering the summation of objective gains over all output variables. Moreover, we extend histogram approximation into multiple output case to speed up the training process. Various experiments on synthetic and real-world datasets verify that GBDT-MO achieves outstanding performance in terms of both accuracy and training speed. Our codes are available on-line.

cs.CV↗

arXiv2019

Pattern formation in a driven Bose-Einstein Condensate

Pattern formation is ubiquitous in nature from morphogenesis and cloud formation to galaxy filamentation. More often than not, patterns arise in a medium driven far from equilibrium due to the interplay of dynamical instability and nonlinear wave mixing. We report, based on momentum and real space pattern recognition, formation of density patterns with two- (D$_2$), four- (D$_4$) and six-fold (D$_6$) symmetries in Bose-Einstein condensates (BECs) with atomic interactions driven at two frequencies. The symmetry of the pattern is controlled by the ratio of the frequencies. The D$_6$ density waves, in particular, arise from a resonant wave mixing process that coherently correlates and enhances the excitations that respect the symmetry.

cond-mat.quant-gas↗

arXiv2019

Adversarial Defense by Suppressing High-frequency Components

Recent works show that deep neural networks trained on image classification dataset bias towards textures. Those models are easily fooled by applying small high-frequency perturbations to clean images. In this paper, we learn robust image classification models by removing high-frequency components. Specifically, we develop a differentiable high-frequency suppression module based on discrete Fourier transform (DFT). Combining with adversarial training, we won the 5th place in the IJCAI-2019 Alibaba Adversarial AI Challenge. Our code is available online.

cs.CV↗

arXiv2019

Recurrent Convolution for Compact and Cost-Adjustable Neural Networks: An Empirical Study

Recurrent convolution (RC) shares the same convolutional kernels and unrolls them multiple steps, which is originally proposed to model time-space signals. We argue that RC can be viewed as a model compression strategy for deep convolutional neural networks. RC reduces the redundancy across layers. However, the performance of an RC network is not satisfactory if we directly unroll the same kernels multiple steps. We propose a simple yet effective variant which improves the RC networks: the batch normalization layers of an RC module are learned independently (not shared) for different unrolling steps. Moreover, we verify that RC can perform cost-adjustable inference which is achieved by varying its unrolling steps. We learn double independent BN layers for cost-adjustable RC networks, i.e. independent w.r.t both the unrolling steps of current cell and upstream cell. We provide insights on why the proposed method works successfully. Experiments on both image classification and image denoise demonstrate the effectiveness of our method.

cs.CV↗

arXiv2018

Quantum Simulation of Coherent Hawking-Unruh Radiation

Exploring quantum phenomena in a curved spacetime is an emerging interdisciplinary area relating many fields in physics such as general relativity, thermodynamics, and quantum information. One famous prediction is the Hawking-Unruh thermal radiation, the manifestation of Minkowski vacuum in an accelerating reference frame. We simulate the radiation by evolving a parametrically driven Bose-Einstein condensate of $\approx 10^5$ atoms, which radiates coherent pairs of atoms with opposite momenta. We observe a matterwave field which follows a Boltzmann distribution for a local observer. The extracted temperature and entropy from the atomic distribution are in agreement with Unruh's predictions. We further observe the long-distance phase coherence and temporal reversibility of emitted matter-waves, hallmarks that distinguish Unruh radiations from classical counterparts. Our results may lead to further insights regarding the nature of the Hawking and Unruh effects and behaviors of quantum physics in a curved spacetime.

physics.atom-ph↗

arXiv2017

Unambiguous nuclear spin detection using engineered quantum sensing sequence

Sensing, localising and identifying individual nuclear spins or frequency components of a signal in the presence of a noisy environments requires the development of robust and selective methods of dynamical decoupling. An important challenge that remains to be addressed in this context are spurious higher order resonances in current dynamical decoupling sequences as they can lead to the misidentification of nuclei or of different frequency components of external signals. Here we overcome this challenge with engineered quantum sensing sequences that achieve both, enhanced robustness and the simultaneous suppression of higher order harmonic resonances. We demonstrate experimentally the principle using a single nitrogen-vacancy center spin sensor which we apply to the unambiguous detection of external protons.

quant-ph↗