arXiv subjects

Jie Ren

Publications and source records attributed to Jie Ren.

At least 271 records · Page 15Linked to original sources

arXiv2018

Explaining Neural Networks Semantically and Quantitatively

This paper presents a method to explain the knowledge encoded in a convolutional neural network (CNN) quantitatively and semantically. The analysis of the specific rationale of each prediction made by the CNN presents a key issue of understanding neural networks, but it is also of significant practical values in certain applications. In this study, we propose to distill knowledge from the CNN into an explainable additive model, so that we can use the explainable model to provide a quantitative explanation for the CNN prediction. We analyze the typical bias-interpreting problem of the explainable model and develop prior losses to guide the learning of the explainable additive model. Experimental results have demonstrated the effectiveness of our method.

cs.CV↗

arXiv2018

Holographic Fermions in Striped Phases

We examine the fermionic response in a holographic model of a low temperature striped phase, working for concreteness with the setup we studied in [Cremonini:2016rbd,Cremonini:2017usb], in which a U(1) symmetry and translational invariance are broken spontaneously at the same time. We include an ionic lattice that breaks translational symmetry explicitly in the UV of the theory. Thus, this construction realizes spontaneous crystallization on top of a background lattice. We solve the Dirac equation for a probe fermion in the associated background geometry using numerical techniques, and explore the interplay between spontaneous and explicit breaking of translations. We note that in our model the breaking of the U(1) symmetry doesn't play a role in the analysis of the fermionic spectral function. We investigate under which conditions a Fermi surface can form and focus in particular on how the ionic lattice affects its structure. When the ionic lattice becomes sufficiently strong the spectral weight peaks broaden, denoting a gradual disappearance of the Fermi surface along the symmetry breaking direction. This phenomenon occurs even in the absence of spontaneously generated stripes. The resulting Fermi surface appears to consist of detached segments reminiscent of Fermi arcs.

hep-th↗

arXiv2018

Quantum-dot circuit-QED thermoelectric diodes and transistors

Recent breakthroughs in quantum-dot circuit-quantum-electrodynamics (circuit-QED) systems are important both from a fundamental perspective and from the point of view of quantum photonic devices. However, understanding the applications of such setups as potential thermoelectric diodes and transistors has been missing. In this paper, via the Keldysh nonequilibrium Green's function approach, we show that cavity-coupled double quantum-dots can serve as excellent quantum thermoelectric diodes and transistors. Using an enhanced perturbation approach based on polaron-transformations, we find non-monotonic dependences of thermoelectric transport properties on the electron-photon interaction. Strong light-matter interaction leads to pronounced rectification effects for both charge and heat, as well as thermal transistor effects in the linear transport regime, which opens up a cutting-edge frontier for quantum thermoelectric devices.

cond-mat.mes-hall↗

arXiv2018

To Compress, or Not to Compress: Characterizing Deep Learning Model Compression for Embedded Inference

The recent advances in deep neural networks (DNNs) make them attractive for embedded systems. However, it can take a long time for DNNs to make an inference on resource-constrained computing devices. Model compression techniques can address the computation issue of deep inference on embedded devices. This technique is highly attractive, as it does not rely on specialized hardware, or computation-offloading that is often infeasible due to privacy concerns or high latency. However, it remains unclear how model compression techniques perform across a wide range of DNNs. To design efficient embedded deep learning solutions, we need to understand their behaviors. This work develops a quantitative approach to characterize model compression techniques on a representative embedded deep learning architecture, the NVIDIA Jetson Tx2. We perform extensive experiments by considering 11 influential neural network architectures from the image classification and the natural language processing domains. We experimentally show that how two mainstream compression techniques, data quantization and pruning, perform on these network architectures and the implications of compression techniques to the model storage size, inference time, energy consumption and performance metrics. We demonstrate that there are opportunities to achieve fast deep inference on embedded systems, but one must carefully choose the compression settings. Our results provide insights on when and how to apply model compression techniques and guidelines for designing efficient embedded deep learning systems.

cs.LG↗

arXiv2018

Linear instability of Poiseuille flows with highly non-ideal fluids

The objective of this work is to investigate linear modal and algebraic instability in Poiseuille flows with fluids close to their vapour-liquid critical point. Close to this critical point, the ideal gas assumption does not hold and large non-ideal fluid behaviours occur. As a representative non-ideal fluid, we consider supercritical carbon dioxide (CO$_2$) at pressure of 80 bar, which is above its critical pressure of 73.9 bar. The Poiseuille flow is characterized by the Reynolds number ($Re=ρ_{w}^{*}u_{r}^{*}h^{*}/μ_{w}^{*}$), the product of Prandtl ($Pr=μ_{w}^{*}C_{pw}^{*}/κ_{w}^{*}$) and Eckert number ($Ec=u_{r}^{*2}/C_{pw}^{*}T_{w}^{*}$), and the wall temperature that in addition to pressure determines the thermodynamic reference condition. For low Eckert numbers, the flow is essentially isothermal and no difference with the well-known stability behaviour of incompressible flows is observed. However, if the Eckert number increases, the viscous heating causes gradients of thermodynamic and transport properties, and non-ideal gas effects become significant. Three regimes of the laminar base flow can be considered, subcritical (temperature in the channel is entirely below its pseudo-critical value), transcritical, and supercritical temperature regime. If compared to the linear stability of an ideal gas Poiseuille flow, we show that the base flow is more unstable in the subcritical regime, inviscid unstable in the transcritical regime, while significantly more stable in the supercritical regime. Following the corresponding states principle, we expect that qualitatively similar results will be obtained for other fluids at equivalent thermodynamic states.

physics.flu-dyn↗

arXiv2018

Seeing Topological Order and Band Inversion in Optical Diatomic Chain

Recent realizations of exotic topological states in condensed matter and cold atoms have advanced the exploration for topological characteristics, such as invariant topological orders and band inversion. Here we construct a 1D optical diatomic system made of split ring resonators with electric-magnetic couplings to implement the topological features. We experimentally measure the dispersion relationship and sub-lattice pseudo-spin vectors by detecting field strength distributions, which determines the accompanying winding number of bulk band as the topological order. The sub-lattice pseudo-spin vector inversion observed at band-edges evidences the band inversion. We further reconfirm the band-inversion-induced topological phase transition by measuring the symmetry exchanges at band-edges. These results have shown the great potential of split ring resonator platform to experimentally explore the advanced topological characteristics, such as high winding numbers or even high Chern numbers in more exotic setups.

physics.optics↗

arXiv2018

Observation of acoustic spin

Unlike optical waves, acoustic waves in fluids are described by scalar pressure fields, and therefore are considered spinless. Here, we demonstrate experimentally the existence of spin in acoustics. In the interference of two acoustic waves propagating perpendicularly to each other, we observed the spin angular momentum in free space as a result of the rotation of local particle velocity. We successfully measured the acoustic spin, and spin induced torque acting on a lossy acoustic meta-atom that results from absorption of the spin angular momentum. The acoustic spin is also observed in the evanescent field of a guided mode traveling along a metamaterial waveguide. We found spin-momentum locking in acoustic waves whose propagation direction is determined by the sign of spin. The observed acoustic spin could open a new door in acoustics and their applications for the control of wave propagation and particle rotation.

physics.class-ph↗

arXiv2018

The new mode of instability in viscous high-speed boundary layer flows

The new mode of instability found by Tunney et al. is studied with viscous stability theory in this article. When the high-speed boundary layer is subject to certain values of favorable pressure gradient and wall heating, a new mode becomes unstable due to the appearance of the streamwise velocity overshoot ($U(y)>U_\infty$) in the base flow. The present study shows that under practical Reynolds numbers, the new mode can hardly co-exist with conventional first mode and Mack's second mode. Due to the requirement for additional wall heating, the new mode may only lead to laminar-turbulent transition under experimental (artificial) conditions.

physics.flu-dyn↗

arXiv2018

Active Learning for Wireless IoT Intrusion Detection

Internet of Things (IoT) is becoming truly ubiquitous in our everyday life, but it also faces unique security challenges. Intrusion detection is critical for the security and safety of a wireless IoT network. This paper discusses the human-in-the-loop active learning approach for wireless intrusion detection. We first present the fundamental challenges against the design of a successful Intrusion Detection System (IDS) for wireless IoT network. We then briefly review the rudimentary concepts of active learning and propose its employment in the diverse applications of wireless intrusion detection. Experimental example is also presented to show the significant performance improvement of the active learning method over traditional supervised learning approach. While machine learning techniques have been widely employed for intrusion detection, the application of human-in-the-loop machine learning that leverages both machine and human intelligence to intrusion detection of IoT is still in its infancy. We hope this article can assist the readers in understanding the key concepts of active learning and spur further research in this area.

cs.CR↗

arXiv2018

Synthesizing and Controlling Helical Indirect Exchange Interactions at Nonequilibrium

We study the nonequilibrium effects of spin and/or electric currents on the helical indirect exchange interactions of local spins that embedded in general open electronic systems. Especially, besides the synthesized anisotropic Heisenberg interactions, we find that the synthetic helical indirect exchange interactions possess two parts: antisymmetric (Dzyaloshinskii-Moriya interaction) and symmetric (Kaplan-Shekhtman-Entin-Wohlman-Aharony interaction), which are all formulated in terms of Keldysh nonequilibrium Green's functions. The presence of either spin-orbit coupling or spin polarized currents alone is able to synthesize and control the antisymmetric Dzyaloshinskii-Moriya exchange interactions, as the same direction as spin splitting. However, the appearance of symmetric Kaplan-Shekhtman-Entin-Wohlman-Aharony interactions requires both, i.e., the spin-orbit coupling and spin polarized currents with different splitting directions. Our results show the detailed scheme of controlling the sign, magnitude, and direction of indirect Dzyaloshinskii-Moriya vectors and Kaplan-Shekhtman-Entin-Wohlman-Aharony interactions at nonequilibrium in open quantum devices.

cond-mat.mes-hall↗

arXiv2018

Identifying viruses from metagenomic data by deep learning

The recent development of metagenomic sequencing makes it possible to sequence microbial genomes including viruses in an environmental sample. Identifying viral sequences from metagenomic data is critical for downstream virus analyses. The existing reference-based and gene homology-based methods are not efficient in identifying unknown viruses or short viral sequences. Here we have developed a reference-free and alignment-free machine learning method, DeepVirFinder, for predicting viral sequences in metagenomic data using deep learning techniques. DeepVirFinder was trained based on a large number of viral sequences discovered before May 2015. Evaluated on the sequences after that date, DeepVirFinder outperformed the state-of-the-art method VirFinder at all contig lengths. Enlarging the training data by adding millions of purified viral sequences from environmental metavirome samples significantly improves the accuracy for predicting under-represented viruses. Applying DeepVirFinder to real human gut metagenomic samples from patients with colorectal carcinoma (CRC) identified 51,138 viral sequences belonging to 175 bins. Ten bins were associated with the cancer status, indicating their potential use for non-invasive diagnosis of CRC. In summary, DeepVirFinder greatly improved the precision and recall rates of viral identification, and it will significantly accelerate the discovery rate of viruses.

q-bio.GN↗

arXiv2018

Probing bino-wino coannihilation dark matter under the neutrino floor at the LHC

In the minimal supersymmetric standard model (MSSM) the bino-wino coannihilation provides a feasible way to accommodate the observed cosmological dark matter (DM) relic density. However, such a scenario usually predicts a very small DM-nucleon scattering cross section that is below the neutrino floor, and can not be tested by DM direct detection experiments. In this work, we investigate the discovery potential of this bino-wino co-annihilation region by searching for the soft dilepton events from the process $pp \to χ^0_2 (\to \ell^+\ell^- χ^0_1) χ^\pm_1+jets$ at the LHC. We find that the mass of the wino-like $χ^0_2$ can be probed up to about 310 (230) GeV at $2σ$ ($5σ$) level for an integrated luminosity ${\cal L}=300$ fb$^{-1}$. In the future HL-LHC with 3000 fb$^{-1}$ luminosity, the corresponding mass limits can be pushed up to 430 (330) GeV.

hep-ph↗

arXiv2018

TeV SUSY dark matter confronted with the current direct and indirect detection data

In the minimal supersymmetric standard model (MSSM) the lightest superparticle (LSP) can be a TeV neutralino (mainly dominated by higgsino or wino) which serves as a dark matter candidate with correct thermal relic density. In this work we confront the 1-2 TeV neutralino dark matter with the latest direct and indirect detections from PandaX and AMS-02/DAMPE. Considering various scenarios with decoupled sfermions, with \textit{A}-mediated annihilation, with squark or stop coannihilation, we find that the parameter space is stringently constrained by the direct detection limits. In the allowed parameter space, the TeV neutralino dark matter annihilation contribution to the anti-proton flux is found to agree with the AMS-02 data while its contribution to eletron/positron flux is too small to cause any visible excess. The current survived parameter space can be mostly covered by the future direct detection experiment LZ7.2T.

hep-ph↗

arXiv2018

Estimating the number of species to attain sufficient representation in a random sample

The statistical problem of using an initial sample to estimate the number of species in a larger sample has found important applications in fields far removed from ecology. Here we address the general problem of estimating the number of species that will be represented by at least a number r of observations in a future sample. The number r indicates species with sufficient observations, which are commonly used as a necessary condition for any robust statistical inference. We derive a procedure to construct consistent estimators that apply universally for a given population: once constructed, they can be evaluated as a simple function of r. Our approach is based on a relation between the number of species represented at least r times and the higher derivatives of the expected number of species discovered per unit of time. Combining this relation with a rational function approximation, we propose nonparametric estimators that are accurate for both large values of r and long-range extrapolations. We further show that our estimators retain asymptotic behaviors that are essential for applications on large-scale datasets. We evaluate the performance of this approach by both simulation and real data applications for inferences of the vocabulary of Shakespeare and Dickens, the topology of a Twitter social network, and molecular diversity in DNA sequencing data.

stat.ME↗

arXiv2018

Quantum phase transitions in spin-1 XXZ chains with rhombic single-ion anisotropy

We explore the fidelity susceptibility and the quantum coherence along with the entanglement entropy in the ground-state of one-dimensional spin-1 XXZ chains with the rhombic single-ion anisotropy. By using the techniques of density matrix renormalization group, effects of the rhombic single-ion anisotropy on a few information theoretical measures are investigated, such as the fidelity susceptibility, the quantum coherence and the entanglement entropy. Their relations with the quantum phase transitions are also analyzed. The phase transitions from the Y-Néel phase to the Large-$E_x$ or the Haldane phase can be well characterized by the fidelity susceptibility. The second-order derivative of the ground-state energy indicates all the transitions are of second order. We also find that the quantum coherence, the entanglement entropy, the Schmidt gap can be used to detect the critical points of quantum phase transitions. Conclusions drawn from these quantum information observables agree well with each other. Finally we provide a ground-state phase diagram as functions of the exchange anisotropy $Δ$ and the rhombic single-ion anisotropy $E$.

cond-mat.str-el↗

arXiv2018

High-order sideband effect with two-level atoms in a shaken optical lattice

The system of cold two-level atoms inside an optical lattice potential has been used to simulate various models encountered in fundamental and condensed-matter physics. When the optical lattice is periodically shaken, some interesting dynamical effects were observed. In this work, by utilizing the ultra-narrow linewidth of the dipole-forbidden transition in $^{87}$Sr atoms loaded in a shaken optical lattice, we reach a new operation regime where the shaking rate is much larger than linewidth of the two-level transition but much smaller than the frequency interval of the harmonic lattice potential, which was not realizable before. Under such unique conditions, an interesting quantum dynamical effect, the high-order sideband effect (HSE), has been experimentally observed. All the results can be well explained by a developed theoretical model with analytical solution. This HSE is governed by different mechanism as the similar phenomenon observed in semiconductors and has unique characteristics.

physics.atom-ph↗

arXiv2018

Simulating Quantum Spin Hall Effect in Topological Lieb Lattice of Linear Circuit

Inspired by the topological insulator circuit proposed and experimentally verified by Jia., et al. \cite{1}, we theoretically realized the topological Lieb lattice, a line centered square lattice with rich topological properties, in a radio-frequency circuit. We open the topological nontrivial band-gap through specific capacitor-inductor network, which resembles adding intrinsic spin orbit coupling term into the tight binding model. Finally, we discuss the extension of the $ϕ=π/2$ phase change of hopping between sites to arbitrary value, and investigate the topological phase transition of the band structure vary with capacitance, thereby paving the way for designing tunable lattices using the presented framework.

cond-mat.mes-hall↗

arXiv2018

Alignment-Free Sequence Analysis and Applications

Genome and metagenome comparisons based on large amounts of next-generation sequencing (NGS) data pose significant challenges for alignment-based approaches due to the huge data size and the relatively short length of the reads. Alignment-free approaches based on the counts of word patterns in NGS data do not depend on the complete genome and are generally computationally efficient. Thus, they contribute significantly to genome and metagenome comparison. Recently, novel statistical approaches have been developed for the comparison of both long and shotgun sequences. These approaches have been applied to many problems including the comparison of gene regulatory regions, genome sequences, metagenomes, binning contigs in metagenomic data, identification of virus-host interactions, and detection of horizontal gene transfers. We provide an updated review of these applications and other related developments of word-count based approaches for alignment-free sequence analysis.

q-bio.QM↗