arXiv subjects

Jun Zhou

Publications and source records attributed to Jun Zhou.

520 records · Page 29Linked to original sources

arXiv2017

Clothing Retrieval with Visual Attention Model

Clothing retrieval is a challenging problem in computer vision. With the advance of Convolutional Neural Networks (CNNs), the accuracy of clothing retrieval has been significantly improved. FashionNet[1], a recent study, proposes to employ a set of artificial features in the form of landmarks for clothing retrieval, which are shown to be helpful for retrieval. However, the landmark detection module is trained with strong supervision which requires considerable efforts to obtain. In this paper, we propose a self-learning Visual Attention Model (VAM) to extract attention maps from clothing images. The VAM is further connected to a global network to form an end-to-end network structure through Impdrop connection which randomly Dropout on the feature maps with the probabilities given by the attention map. Extensive experiments on several widely used benchmark clothing retrieval data sets have demonstrated the promise of the proposed method. We also show that compared to the trivial Product connection, the Impdrop connection makes the network structure more robust when training sets of limited size are used.

cs.CV↗

arXiv2017

La interstitial defect-induced insulator-metal transition in oxide heterostructures LaAlO3/SrTiO3

Perovskite oxide interfaces have attracted tremendous research interest for their fundamental physics and promising all-oxide electronics applications. Here, based on first-principles calculations, we propose a novel surface La interstitial promoted interface insulator-metal transition in LaAlO3/SrTiO3 (110). Compared with surface oxygen vacancies, which play a determining role on the insulator-metal transition of LaAlO3/SrTiO3 (001) interfaces, we find that surface La interstitials can be more experimentally realistic and accessible for manipulation and more stable in ambient atmospheric environment. Interestingly, these surface La interstitials also induce significant spin-splitting states with Ti dyz/dxz character at conducting LaAlO3/SrTiO3 (110) interface. On the other hand, for insulating LaAlO3/SrTiO3 (110) (<4 unit cells LaAlO3 thickness), a distortion between La (Al) and O atoms is found at the LaAlO3 side, partially compensating the polarization divergence. Our results reveal the origin of metal-insulator transition in LaAlO3/SrTiO3 (110) heterostructures, and also shed light on the manipulation of the superior properties of LaAlO3/SrTiO3 (110) for new possibilities of electronic and magnetic applications.

cond-mat.mtrl-sci↗

arXiv2017

Small-footprint Keyword Spotting Using Deep Neural Network and Connectionist Temporal Classifier

Mainly for the sake of solving the lack of keyword-specific data, we propose one Keyword Spotting (KWS) system using Deep Neural Network (DNN) and Connectionist Temporal Classifier (CTC) on power-constrained small-footprint mobile devices, taking full advantage of general corpus from continuous speech recognition which is of great amount. DNN is to directly predict the posterior of phoneme units of any personally customized key-phrase, and CTC to produce a confidence score of the given phoneme sequence as responsive decision-making mechanism. The CTC-KWS has competitive performance in comparison with purely DNN based keyword specific KWS, but not increasing any computational complexity.

cs.CL↗

arXiv2017

Attributing Hacks

In this paper we describe an algorithm for estimating the provenance of hacks on websites. That is, given properties of sites and the temporal occurrence of attacks, we are able to attribute individual attacks to joint causes and vulnerabilities, as well as estimating the evolution of these vulnerabilities over time. Specifically, we use hazard regression with a time-varying additive hazard function parameterized in a generalized linear form. The activation coefficients on each feature are continuous-time functions over time. We formulate the problem of learning these functions as a constrained variational maximum likelihood estimation problem with total variation penalty and show that the optimal solution is a 0th order spline (a piecewise constant function) with a finite number of known knots. This allows the inference problem to be solved efficiently and at scale by solving a finite dimensional optimization problem. Extensive experiments on real data sets show that our method significantly outperforms Cox's proportional hazard model. We also conduct a case study and verify that the fitted functions are indeed recovering vulnerable features and real-life events such as the release of code to exploit these features in hacker blogs.

cs.LG↗

arXiv2017

Hopping Processes Explain T-linear Rise of Thermal Conductivity in Thermoelectric Clathrates above the Plateau

Type-I clathrate compounds with off-center guest ions realize the phonon-glass electron-crystal concept by exhibiting almost identical lattice thermal conductivities $κ_{\rm L}$ to those observed in network-forming glasses. This is in contrast with type-I clathrates with on-center guest ions showing $κ_{\rm L}$ of conventional crystallines. Glasslike $κ_{\rm L}$ stems from the peculiar THz frequency dynamics in off-center type-I clathrates where there exist three kinds of modes classified into extended(EX), weakly(WL) and strongly localized(SL) modes as demonstrated by Liu $et.\,al.$, Phys. Rev. B $\textbf{93}$, 214305(2016). Our calculated results based on the hopping mechanism of SL modes via anharmonic interactions show fairly good agreement with observed ${\it T}$-linear rise of $κ_{\rm L}$ above the plateau. We emphasize that both the magnitude and the temperature dependence are in accord with the experimental data of off-center type-I clathrates.

cond-mat.other↗

arXiv2017

Identifying Designs from Incomplete, Fragmented Cultural Heritage Objects by Curve-Pattern Matching

Study of cultural-heritage objects with embellished realistic and abstract designs made up of connected and intertwined curves crosscuts a number of related disciplines, including archaeology, art history, and heritage management. However, many objects, such as pottery sherds found in the archaeological record, are fragmentary, making the underlying complete designs unknowable at the scale of the sherd fragment. The challenge to reconstruct and study complete designs is stymied because 1) most fragmentary cultural-heritage objects contain only a small portion of the underlying full design, 2) in the case of a stamping application, the same design may be applied multiple times with spatial overlap on one object, and 3) curve patterns detected on an object are usually incomplete and noisy. As a result, classical curve-pattern matching algorithms, such as Chamfer matching, may perform poorly in identifying the underlying design. In this paper, we develop a new partial-to-global curve matching algorithm to address these challenges and better identify the full design from a fragmented cultural heritage object. Specifically, we develop the algorithm to identify the designs of the carved wooden paddles of the Southeastern Woodlands from unearthed pottery sherds. A set of pottery sherds from the Snow Collection, curated at Georgia Southern University, are used to test the proposed algorithm, with promising results.

cs.CV↗

arXiv2016

On the Sampling Strategy for Evaluation of Spectral-spatial Methods in Hyperspectral Image Classification

Spectral-spatial processing has been increasingly explored in remote sensing hyperspectral image classification. While extensive studies have focused on developing methods to improve the classification accuracy, experimental setting and design for method evaluation have drawn little attention. In the scope of supervised classification, we find that traditional experimental designs for spectral processing are often improperly used in the spectral-spatial processing context, leading to unfair or biased performance evaluation. This is especially the case when training and testing samples are randomly drawn from the same image - a practice that has been commonly adopted in the experiments. Under such setting, the dependence caused by overlap between the training and testing samples may be artificially enhanced by some spatial information processing methods such as spatial filtering and morphological operation. Such interaction between training and testing sets has violated data independence assumption that is abided by supervised learning theory and performance evaluation mechanism. Therefore, the widely adopted pixel-based random sampling strategy is not always suitable to evaluate spectral-spatial classification algorithms because it is difficult to determine whether the improvement of classification accuracy is caused by incorporating spatial information into classifier or by increasing the overlap between training and testing samples. To partially solve this problem, we propose a novel controlled random sampling strategy for spectral-spatial methods. It can greatly reduce the overlap between training and testing samples and provides more objective and accurate evaluation.

cs.CV↗

arXiv2016

Spin-dependent Seebeck Effect in Aharonov-Bohm Rings with Rashba and Dresselhaus Spin-orbit Interactions

We theoretically investigate the spin-dependent Seebeck effect in an Aharonov-Bohm mesoscopic ring in the presence of both Rashba and Dresselhaus spin-orbit interactions under magnetic flux perpendicular to the ring. We apply the Green's function method to calculate the spin Seebeck coefficient employing the tight-binding Hamiltonian. It is found that the spin Seebeck coefficient is proportional to the slope of the energy-dependent transmission coefficients. We study the strong dependence of spin Seebeck coefficient on the Fermi energy, magnetic flux, strength of spin-orbit coupling, and temperature. Maximum spin Seebeck coefficients can be obtained when the strengths of Rashba and Dresselhaus spin-orbit couplings are slightly different. The spin Seebeck coefficient can be reduced by increasing temperature and disorder.

cond-mat.mes-hall↗

arXiv2016

Interfacial thermal conductance across metal-insulator/semiconductor interfaces due to surface states

We point out that the effective channel for the interfacial thermal conductance, the inverse of Kapitza resistance, of metal-insulator/semiconductor interfaces is governed by the electron-phonon interaction mediated by the surface states allowed in a thin region near the interface. Our detailed calculations demonstrate that the interfacial thermal conductance across Pb/Pt/Al/Au-diamond interfaces are only slightly different among these metals, and reproduce well the experimental results of the interfacial thermal conductance across metal-diamond interfaces observed by Stoner et al. [Phys. Rev. Lett. 68, 1563 (1992)] and most recently by Hohensee et al. [Nature Commun. 6, 6578 (2015)].

cond-mat.mes-hall↗

arXiv2015

The interplay of electronic reconstructions, lattice distortions, and surface oxygen vacancies in insulator-metal transition of LaAlO$_{3}$/SrTiO$_{3}$

The mechanism responsible for the extraordinary interface conductivity of LaAlO$_{3}$ on SrTiO$_{3}$ and its insulator-metal transition remains controversial. Here, using density functional theory calculations, we establish a comprehensive and coherent picture that the interplay of electronic reconstructions, lattice distortions, and surface oxygen vacancies fully compensates the polarization potential divergence in LaAlO$_{3}$/SrTiO$_{3}$, explaining naturally the experimental observations under different conditions. While lattice distortions and a charge redistribution between LaO and AlO$_2$ sub-layers play a dominant role in insulating state, a spontaneous appearance of 1/4 oxygen vacancies per AlO$_{2}$ sub-layer at the LaAlO$_{3}$ surface accompanied by 0.5$e^{-}$ charge-transfer into the interface is responsible for interface conductivity and the discontinuous transition in LaAlO$_{3}$/SrTiO$_{3}$. Our model also explain properties of LaAlO$_{3}$/SrTiO$_{3}$ prepared with different growth conditions.

cond-mat.str-el↗

arXiv2014

On Transition Metal Catalyzed Reduction of N-nitrosodimethlamine

This report provides a critical review on "Metal-Catalyzed Reduction of N-Nitrosodimethylamine with Hydrogen in Water", by Davie et al. N-nitrosodimethlamine (NDMA) is a contaminant in drinking and ground water which is difficult to remove by conventional physical methods, such as air stripping. Based on the reported robust capability of metal based powder shaped catalysts in hydrogen reduction, several monometallic and bimetallic catalyst are studied in this paper on the reduction of NDMA with hydrogen. Two kinds of kinetics, metal weight normalized and surface area normalized, are compared between each catalyst in terms of pseudo-first order reaction rate. Palladium, copper enhanced palladium and nickel are found to be very efficient in NDMA reduction, with half-lives on the order of hours per 10 mg/l catalyst metal. Preliminary LC-MS data and carbon balance showed no intermediates. Finally, a simple hydrogen and NMDA surface activated reaction mechanism is proposed by the author for palladium and nickel.

physics.chem-ph↗

arXiv2014

Thermal Boundary Conductance Across Metal-Nonmetal Interfaces: Effects of Electron-Phonon Coupling both in Metal and at Interface

We theoretically investigate the thermal boundary conductance across metal-nonmetal interfaces in the presence of the electron-phonon coupling not only in metal but also at interface. The thermal energy can be transferred from metal to nonmetal via three channels: (1) the phonon-phonon coupling at interface; (2) the electron-phonon coupling at interface; and (3) the electron-phonon coupling within metal and then subsequently the phonon-phonon coupling at interface. We find that these three channels can be described by an equivalent series-parallel thermal resistor network, based on which we derive out the analytic expression of the thermal boundary conductance. We then exemplify different contributions from each channel to the thermal boundary conductance in three typical interfaces: Pb-diamond, Ti-diamond, and TiN-MgO. Our results reveal that the competition among above channels determines the thermal boundary conductance.

cond-mat.mtrl-sci↗

arXiv2014

Spin Seebeck Effect in Asymmetric Four-Terminal Systems with Rashba Spin-Orbit Coupling

We propose a new type of the spin Seebeck effect (SSE) emerging from the Rashba spin-orbit coupling in asymmetric four-terminal electron systems. This system generates spin currents or spin voltages along the longitudinal direction parallel to the temperature gradient in the absence of magnetic fields. The remarkable result arises from the breaking of reflection symmetry along the transverse direction. In the meantime, the SSE along the transverse direction, so-called the spin Nernst effect, with spin currents or spin voltages perpendicular to the temperature gradient can be simultaneously realized in our system. We further find that it is possible to use the temperature differences between four leads to tune the spin Seebeck coefficients.

cond-mat.mes-hall↗

arXiv2014

An Electrohydrodynamics Model for Non-equilibrium Electron and Phonon Transport in Metal Films after Ultra-short Pulse Laser Heating

The electrons and phonons in metal films after ultra-short pulse laser heating are in highly non-equilibrium states not only between the electron sub-system and the phonon sub-system but also within the electron sub-system. An electrohydrodynamics model consisting of the balance equations of electron density, energy density of electrons, and energy density of phonons is derived from the coupled non-equilibrium electron and phonon Boltzmann transport equations to study the nonlinear transport phenomena, such as the electron density fluctuation and the transient electrical current in metal films, after ultra-short pulse laser heating. The time-dependent temperature distributions is calculated by the coupled electron and phonon Boltzmann transport equations, the electrohydrodynamics model derived in this work, and the two-temperature model for different laser pulse durations, film thicknesses, and laser fluences. We find that the two-temperature model overestimates the electron temperature at the frontsurface of the film and underestimates the damage threshold when the nonlinear thermal transport of electrons is important. The electrohydrodynamics model proposedin this work could be a more accurate prediction tool to study the non-equilibrium electron phonon transport process than the two-temperature model and it is much easier to be solved than the coupled electron and phonon Boltzmann transport equations.

cond-mat.mtrl-sci↗

arXiv2014

Inhomogeneous Thermal Conductivity Enhances Thermoelectric Cooling

We theoretically investigate the enhancement of thermoelectric cooling performance in thermoelectric devices made of materials with inhomogeneous thermal conductivity, beyond the usual practice of enhancing thermoelectric figure of merit ZT. The dissipation of Joule heat in such thermoelectric devices is asymmetric which can give rise to better thermoelectric cooling performance. Although the thermoelectric figure of merit and the coefficient-of-performance are only slightly enhanced, both the maximum cooling power and the maximum cooling temperature difference can be enhanced significantly. This finding can be used to increase the heat absorption at the cold end. The asymmetric dissipation of Joule heat also leads to thermal rectification.

cond-mat.mes-hall↗

arXiv2014

Wave-packet rectification in nonlinear electronic systems: A tunable Aharonov-Bohm diode

Rectification of electron wave-packets propagating along a quasi-one dimensional chain is commonly achieved via the simultaneous action of nonlinearity and longitudinal asymmetry, both confined to a limited portion of the chain termed wave diode. However, it is conceivable that, in the presence of an external magnetic field, spatial asymmetry perpendicular to the direction of propagation suffices to ensure rectification. This is the case of a nonlinear ring-shaped lattice with different upper and lower halves (diode), which is attached to two elastic chains (leads). The resulting device is mirror symmetric with respect to the ring vertical axis, but mirror asymmetric with respect to the chain direction. Wave propagation along the two diode paths can be modeled for simplicity by a discrete Schrödinger equation with cubic nonlinearities. Numerical simulations demonstrate that, thanks to the Aharonov-Bohm effect, such a diode can be operated by tuning the magnetic flux across the ring.

physics.comp-ph↗