Searcharxiv⌕ Search

arXiv subjects

Rui Lin

Publications and source records attributed to Rui Lin.

At least 55 records · Page 3Linked to original sources
arXiv2022

Observing dynamical currents in a non-Hermitian momentum lattice

We report on the experimental realization and detection of dynamical currents in a spin-textured lattice in momentum space. Collective tunneling is implemented via cavity-assisted Raman scattering of photons by a spinor Bose-Einstein condensate into an optical cavity. The photon field inducing the tunneling processes is subject to cavity dissipation, resulting in effective directional dynamics in a non-Hermitian setting. We observe that the individual tunneling events are superradiant in nature and locally resolve them in the lattice by performing real-time, frequency-resolved measurements of the leaking cavity field. The results can be extended to a regime exhibiting a cascade of currents and simultaneous coherences between multiple lattice sites, where numerical simulations provide further understanding of the dynamics. Our observations showcase dynamical tunneling in momentum-space lattices and provide prospects to realize dynamical gauge fields in driven-dissipative settings.

cond-mat.quant-gas↗
arXiv2022

Coarse to Fine: Image Restoration Boosted by Multi-Scale Low-Rank Tensor Completion

Existing low-rank tensor completion (LRTC) approaches aim at restoring a partially observed tensor by imposing a global low-rank constraint on the underlying completed tensor. However, such a global rank assumption suffers the trade-off between restoring the originally details-lacking parts and neglecting the potentially complex objects, making the completion performance unsatisfactory on both sides. To address this problem, we propose a novel and practical strategy for image restoration that restores the partially observed tensor in a coarse-to-fine (C2F) manner, which gets rid of such trade-off by searching proper local ranks for both low- and high-rank parts. Extensive experiments are conducted to demonstrate the superiority of the proposed C2F scheme. The codes are available at: https://github.com/RuiLin0212/C2FLRTC.

cs.CV↗
arXiv2022

Deformable Butterfly: A Highly Structured and Sparse Linear Transform

We introduce a new kind of linear transform named Deformable Butterfly (DeBut) that generalizes the conventional butterfly matrices and can be adapted to various input-output dimensions. It inherits the fine-to-coarse-grained learnable hierarchy of traditional butterflies and when deployed to neural networks, the prominent structures and sparsity in a DeBut layer constitutes a new way for network compression. We apply DeBut as a drop-in replacement of standard fully connected and convolutional layers, and demonstrate its superiority in homogenizing a neural network and rendering it favorable properties such as light weight and low inference complexity, without compromising accuracy. The natural complexity-accuracy tradeoff arising from the myriad deformations of a DeBut layer also opens up new rooms for analytical and practical research. The codes and Appendix are publicly available at: https://github.com/ruilin0212/DeBut.

cs.CV↗
arXiv2022

What Do Adversarially trained Neural Networks Focus: A Fourier Domain-based Study

Although many fields have witnessed the superior performance brought about by deep learning, the robustness of neural networks remains an open issue. Specifically, a small adversarial perturbation on the input may cause the model to produce a completely different output. Such poor robustness implies many potential hazards, especially in security-critical applications, e.g., autonomous driving and mobile robotics. This work studies what information the adversarially trained model focuses on. Empirically, we notice that the differences between the clean and adversarial data are mainly distributed in the low-frequency region. We then find that an adversarially-trained model is more robust than its naturally-trained counterpart due to the reason that the former pays more attention to learning the dominant information in low-frequency components. In addition, we consider two common ways to improve model robustness, namely, by data augmentation and by using stronger network architectures, and understand these techniques from a frequency-domain perspective. We are hopeful this work can shed light on the design of more robust neural networks.

cs.CV↗
arXiv2022

Dissipation-engineered family of nearly dark states in many-body cavity-atom systems

Three-level atomic systems coupled to light have the capacity to host dark states. We study a system of V-shaped three-level atoms coherently coupled to the two quadratures of a dissipative cavity. The interplay between the atomic level structure and dissipation makes the phase diagram of the open system drastically different from the closed one. In particular, it leads to the stabilization of a continuous family of dark and nearly dark excited many-body states with inverted atomic populations as the steady states. The multistability of these states can be probed via their distinct fluctuations and excitation spectra, as well as the system's Liouvillian dynamics which are highly sensitive to ramp protocols. Our model can be implemented experimentally by encoding the two higher-energy modes in orthogonal density-modulated states in a bosonic quantum gas. This implementation offers prospects for potential applications like the realization of quantum optical random walks and microscopy with subwavelength spatial resolution.

cond-mat.quant-gas↗
arXiv2021

Mott transition in a cavity-boson system: A quantitative comparison between theory and experiment

The competition between short-range and cavity-mediated infinite-range interactions in a cavity-boson system leads to the existence of a superfluid phase and a Mott-insulator phase within the self-organized regime. In this work, we quantitatively compare the steady-state phase boundaries of this transition measured in experiments and simulated using the Multiconfigurational Time-Dependent Hartree Method for Indistinguishable Particles. To make the problem computationally feasible, we represent the full system by the exact many-body wave function of a two-dimensional four-well potential. We argue that the validity of this representation comes from the nature of both the cavity-atomic system and the Bose-Hubbard physics. Additionally we show that the chosen representation only induces small systematic errors, and that the experimentally measured and theoretically predicted phase boundaries agree reasonably. We thus demonstrate a new approach for the quantitative numerical determination of the superfluid--Mott-insulator phase boundary.

cond-mat.quant-gas↗
arXiv2021

EZCrop: Energy-Zoned Channels for Robust Output Pruning

Recent results have revealed an interesting observation in a trained convolutional neural network (CNN), namely, the rank of a feature map channel matrix remains surprisingly constant despite the input images. This has led to an effective rank-based channel pruning algorithm, yet the constant rank phenomenon remains mysterious and unexplained. This work aims at demystifying and interpreting such rank behavior from a frequency-domain perspective, which as a bonus suggests an extremely efficient Fast Fourier Transform (FFT)-based metric for measuring channel importance without explicitly computing its rank. We achieve remarkable CNN channel pruning based on this analytically sound and computationally efficient metric and adopt it for repetitive pruning to demonstrate robustness via our scheme named Energy-Zoned Channels for Robust Output Pruning (EZCrop), which shows consistently better results than other state-of-the-art channel pruning methods.

eess.IV↗
arXiv2021

Exploiting Elasticity in Tensor Ranks for Compressing Neural Networks

Elasticities in depth, width, kernel size and resolution have been explored in compressing deep neural networks (DNNs). Recognizing that the kernels in a convolutional neural network (CNN) are 4-way tensors, we further exploit a new elasticity dimension along the input-output channels. Specifically, a novel nuclear-norm rank minimization factorization (NRMF) approach is proposed to dynamically and globally search for the reduced tensor ranks during training. Correlation between tensor ranks across multiple layers is revealed, and a graceful tradeoff between model size and accuracy is obtained. Experiments then show the superiority of NRMF over the previous non-elastic variational Bayesian matrix factorization (VBMF) scheme.

cs.LG↗
arXiv2021

FAT: Learning Low-Bitwidth Parametric Representation via Frequency-Aware Transformation

Learning convolutional neural networks (CNNs) with low bitwidth is challenging because performance may drop significantly after quantization. Prior arts often discretize the network weights by carefully tuning hyper-parameters of quantization (e.g. non-uniform stepsize and layer-wise bitwidths), which are complicated and sub-optimal because the full-precision and low-precision models have a large discrepancy. This work presents a novel quantization pipeline, Frequency-Aware Transformation (FAT), which has several appealing benefits. (1) Rather than designing complicated quantizers like existing works, FAT learns to transform network weights in the frequency domain before quantization, making them more amenable to training in low bitwidth. (2) With FAT, CNNs can be easily trained in low precision using simple standard quantizers without tedious hyper-parameter tuning. Theoretical analysis shows that FAT improves both uniform and non-uniform quantizers. (3) FAT can be easily plugged into many CNN architectures. When training ResNet-18 and MobileNet-V2 in 4 bits, FAT plus a simple rounding operation already achieves 70.5% and 69.2% top-1 accuracy on ImageNet without bells and whistles, outperforming recent state-of-the-art by reducing 54.9X and 45.7X computations against full-precision models. We hope FAT provides a novel perspective for model quantization. Code is available at \url{https://github.com/ChaofanTao/FAT_Quantization}.

cs.CV↗
arXiv2020

Optimized Observable Readout from Single-shot Images of Ultracold Atoms via Machine Learning

Single-shot images are the standard readout of experiments with ultracold atoms -- the tarnished looking glass into their many-body physics. The efficient extraction of observables from single-shot images is thus crucial. Here, we demonstrate how artificial neural networks can optimize this extraction. In contrast to standard averaging approaches, machine learning allows both one- and two-particle densities to be accurately obtained from a drastically reduced number of single-shot images. Quantum fluctuations and correlations are directly harnessed to obtain physical observables for bosons in a tilted double-well potential at an unprecedented accuracy. Strikingly, machine learning also enables a reliable extraction of momentum-space observables from real-space single-shot images and vice versa. This obviates the need for a reconfiguration of the experimental setup between in-situ and time-of-flight imaging, thus potentially granting an outstanding reduction in resources.

cond-mat.quant-gas↗
arXiv2020

Pathway to chaos through hierarchical superfluidity in a cavity-BEC system

We explore the role of atomic correlations in a harmonically trapped Bose-Einstein condensate coupled to a dissipative cavity, where both the atoms and the cavity are blue detuned from the external pumping laser. Using a genuine many-body approach that goes beyond mean-field, we extract density distributions and many-body correlations to unveil a pathway to chaos at large pump power through a hierarchical self-organization of the atoms, where the atoms transition from a single-well optical lattice to a double-well optical lattice. Correlated states of the atoms emerge and are characterized by local superfluid correlations in phases which are globally superfluid or Mott insulating. Local superfluid-Mott transitions are precluded by a dynamical instability to chaos which occurs via quasiperiodic attractors. Our results explain the mechanism behind the dynamical instabilities observed in experiments.

cond-mat.quant-gas↗
arXiv2020

HOTCAKE: Higher Order Tucker Articulated Kernels for Deeper CNN Compression

The emerging edge computing has promoted immense interests in compacting a neural network without sacrificing much accuracy. In this regard, low-rank tensor decomposition constitutes a powerful tool to compress convolutional neural networks (CNNs) by decomposing the 4-way kernel tensor into multi-stage smaller ones. Building on top of Tucker-2 decomposition, we propose a generalized Higher Order Tucker Articulated Kernels (HOTCAKE) scheme comprising four steps: input channel decomposition, guided Tucker rank selection, higher order Tucker decomposition and fine-tuning. By subjecting each CONV layer to HOTCAKE, a highly compressed CNN model with graceful accuracy trade-off is obtained. Experiments show HOTCAKE can compress even pre-compressed models and produce state-of-the-art lightweight networks.

cs.LG↗
arXiv2020

MCTDH-X: The multiconfigurational time-dependent Hartree method for indistinguishable particles software

We introduce and describe the multiconfigurational time-depenent Hartree for indistinguishable particles (MCTDH-X) software. This powerful tool allows the investigation of ground state properties and dynamics of interacting quantum many-body systems in different spatial dimensions. The MCTDH-X software is a set of programs and scripts to compute, analyze, and visualize solutions for the time-dependent and time-independent many-body Schrödinger equation for indistinguishable quantum particles. As the MCTDH-X software represents a general solver for the Schrödinger equation, it is applicable to a wide range of problems in the fields of atomic, optical, molecular physics as well as condensed matter systems. In particular, it can be used to study light-matter interactions, correlated dynamics of electrons, as well as some aspects related to quantum information and computing. The MCTDH-X software solves a set of non-linear coupled working equations based on the application of the variational principle to the Schrödinger equation. These equations are obtained by using an ansatz for the many-body wavefunction that is a time-dependent expansion in a set of time-dependent many-body basis states. The time-dependence of the basis set enables MCTDH-X to deal with quantum dynamics at a superior accuracy as compared to, for instance, exact diagonalization approaches. Herein, we give an introduction to the MCTDH-X software via an easy-to-follow tutorial with a focus on accessibility. We use the double well to illustrate the fermionization of bosonic particles, the crystallization of fermionic particles, characteristics of the superfluid and Mott-insulator quantum phases in Hubbard models, and even dynamical quantum phase transitions. Our tutorial guides the potential user to apply the MCTDH-X software also to more complex systems.

cond-mat.quant-gas↗
arXiv2019

Spiral order from orientationally correlated random bonds in classical XY models

We discuss the stability of ferromagnetic long-range order in three-dimensional classical XY ferromagnets upon substitution of a small subset of equally oriented bonds by impurity bonds, on which the ferromagnetic exchange J_perp > 0 is replaced by a strong antiferromagnetic coupling J_imp < 0. In the presence of a single impurity bond, once the absolute value of the frustrating coupling J_imp < 0 exceeds a threshold J_c > 0, the ground state becomes two-fold degenerate, corresponding to either clockwise or anticlockwise canting of the spins in the vicinity of the impurity bond. In the presence of a small concentration of impurity bonds, the effective low-energy Hamiltonian is that of Ising variables encoding the sense of rotation of the local canting around the impurities. Those degrees of freedom interact through a dipolar interaction mediated by spin waves. A ferromagnetic Ising ground state indicates the instability of the XY ferromagnet towards a spiral state with a wave vector proportional to the concentration of impurity bonds. To analyze under which circumstances such a ground state arises, we study first impurities forming superlattices. For a subclass of those, we can rigorously establish the existence of spiral order. For another class of superlattices, the Ising variables order ferromagnetically in planes perpendicular to the orientation of impurity bonds, but antiferromagnetically parallel to it, which results in a fan-like XY ground state. Second, we consider the case when the impurity bonds are randomly distributed on the three-dimensional host lattice according to a Poisson process. We show the phenomenon of spiral order by disorder with an ordering wave vector proportional to the impurity concentration. The analytical predictions are confirmed by Monte Carlo simulations and are relevant for magnetic materials such as YBaCuFeO_5.

cond-mat.str-el↗
arXiv2019

MiSC: Mixed Strategies Crowdsourcing

Popular crowdsourcing techniques mostly focus on evaluating workers' labeling quality before adjusting their weights during label aggregation. Recently, another cohort of models regard crowdsourced annotations as incomplete tensors and recover unfilled labels by tensor completion. However, mixed strategies of the two methodologies have never been comprehensively investigated, leaving them as rather independent approaches. In this work, we propose $\textit{MiSC}$ ($\textbf{Mi}$xed $\textbf{S}$trategies $\textbf{C}$rowdsourcing), a versatile framework integrating arbitrary conventional crowdsourcing and tensor completion techniques. In particular, we propose a novel iterative Tucker label aggregation algorithm that outperforms state-of-the-art methods in extensive experiments.

cs.HC↗
arXiv2018

Superfluid--Mott insulator transition of ultracold superradiant bosons in a cavity

We investigate harmonically-trapped, laser-pumped bosons with infinite-range interactions induced by a dissipative high-finesse red-detuned optical cavity with numerical and analytical methods. We obtain multiple cavity and atomic observables as well as the full phase diagram of the system using the multiconfigurational time-dependent Hartree method for indistinguishable particles (MCTDH-X) approach. Besides the transition from an unorganized normal phase to a superradiant phase where atoms self-organize, we focus on an in-depth investigation of the self-organized superfluid to self-organized Mott insulator phase transition in the superradiant phase as a function of the cavity-atom coupling. The numerical results are substantiated by an analytical study of an effective Bose-Hubbard model. We numerically analyze cavity fluctuations and emergent strong correlations between atoms in the many-body state across the Mott transition via the atomic density distributions and Glauber correlation functions. Unexpectedly, the weak harmonic trap leads to features like a lattice switching between the two symmetry-broken $\mathbb{Z}_2$ configurations of the untrapped system and a reentrance of superfluidity in the Mott insulating phase. Our analytical considerations quantitatively explain the numerically observed correlation features.

cond-mat.quant-gas↗
arXiv2018

High-speed PAM4-based Optical SDM Interconnects with Directly Modulated Long-wavelength VCSEL

This paper reports the demonstration of high-speed PAM-4 transmission using a 1.5-μm single-mode vertical cavity surface emitting laser (SM-VCSEL) over multicore fiber with 7 cores over different distances. We have successfully generated up to 70 Gbaud 4-level pulse amplitude modulation (PAM-4) signals with a VCSEL in optical back-to-back, and transmitted 50 Gbaud PAM-4 signals over both 1-km dispersion-uncompensated and 10-km dispersion-compensated in each core, enabling a total data throughput of 700 Gbps over the 7-core fiber. Moreover, 56 Gbaud PAM-4 over 1-km has also been shown, whereby unfortunately not all cores provide the required 3.8 $\times$ 10 $^{-3}$ bit error rate (BER) for the 7% overhead-hard decision forward error correction (7% OH HDFEC). The limited bandwidth of the VCSEL and the adverse chromatic dispersion of the fiber are suppressed with pre-equalization based on accurate end-to-end channel characterizations. With a digital post-equalization, BER performance below the 7% OH-HDFEC limit is achieved over all cores. The demonstrated results show a great potential to realize high-capacity and compact short-reach optical interconnects for data centers.

eess.SP↗