arXiv subjects

Gang He

Publications and source records attributed to Gang He.

27 records · Page 2Linked to original sources

arXiv2021

Construct Informative Triplet with Two-stage Hard-sample Generation

In this paper, we propose a robust sample generation scheme to construct informative triplets. The proposed hard sample generation is a two-stage synthesis framework that produces hard samples through effective positive and negative sample generators in two stages, respectively. The first stage stretches the anchor-positive pairs with piecewise linear manipulation and enhances the quality of generated samples by skillfully designing a conditional generative adversarial network to lower the risk of mode collapse. The second stage utilizes an adaptive reverse metric constraint to generate the final hard samples. Extensive experiments on several benchmark datasets verify that our method achieves superior performance than the existing hard-sample generation algorithms. Besides, we also find that our proposed hard sample generation method combining the existing triplet mining strategies can further boost the deep metric learning performance.

cs.CV↗

arXiv2021

Alloy Engineering of Polar (Si,Ge)2N2O System for Controllable Second Harmonic Performance

Although silicon oxynitrides are important semiconductors for many practical applications, their potential second-order nonlinear optical (NLO) applications, regardless of balanced or controllable performance, have never been systemically explored. Using the first-principles calculations, in this article, we discover that the sinoite (i.e., typical silicon oxynitride Si2N2O) can simultaneously exhibit wide optical bandgap, strong second-harmonic generation (SHG) effect, and large birefringence, which are further confirmed by our preliminary experimental data. Importantly, we propose that alloying engineering can be further applied to control the balanced NLO properties in the Si2N2O system. Combining first-principles calculations and cluster expansion theory, we demonstrate that alloying Ge into Si2N2O can easily form low formation energy Si2(1-x)Ge2xN2O alloys, which can in turn achieve controllable phase-matching harmonic output with high SHG efficiency at different energy ranges. Therefore, alloy engineering could provide a unique approach to effectively control the balanced NLO performance of Si2(1-x)Ge2xN2O, making this polar alloy system holding potential applications in tunable laser frequency conversion and controllable all-optical devices.

cond-mat.mtrl-sci↗

arXiv2021

RR-DnCNN v2.0: Enhanced Restoration-Reconstruction Deep Neural Network for Down-Sampling Based Video Coding

Integrating deep learning techniques into the video coding framework gains significant improvement compared to the standard compression techniques, especially applying super-resolution (up-sampling) to down-sampling based video coding as post-processing. However, besides up-sampling degradation, the various artifacts brought from compression make super-resolution problem more difficult to solve. The straightforward solution is to integrate the artifact removal techniques before super-resolution. However, some helpful features may be removed together, degrading the super-resolution performance. To address this problem, we proposed an end-to-end restoration-reconstruction deep neural network (RR-DnCNN) using the degradation-aware technique, which entirely solves degradation from compression and sub-sampling. Besides, we proved that the compression degradation produced by Random Access configuration is rich enough to cover other degradation types, such as Low Delay P and All Intra, for training. Since the straightforward network RR-DnCNN with many layers as a chain has poor learning capability suffering from the gradient vanishing problem, we redesign the network architecture to let reconstruction leverages the captured features from restoration using up-sampling skip connections. Our novel architecture is called restoration-reconstruction u-shaped deep neural network (RR-DnCNN v2.0). As a result, our RR-DnCNN v2.0 outperforms the previous works and can attain 17.02% BD-rate reduction on UHD resolution for all-intra anchored by the standard H.265/HEVC. The source code is available at https://minhmanho.github.io/rrdncnn/.

eess.IV↗

arXiv2016

Critical noise of majority-vote model on complex networks

The majority-vote model with noise is one of the simplest nonequilibrium statistical model that has been extensively studied in the context of complex networks. However, the relationship between the critical noise where the order-disorder phase transition takes place and the topology of the underlying networks is still lacking. In the paper, we use the heterogeneous mean-field theory to derive the rate equation for governing the model's dynamics that can analytically determine the critical noise $f_c$ in the limit of infinite network size $N\rightarrow \infty$. The result shows that $f_c$ depends on the ratio of ${\left\langle k \right\rangle }$ to ${\left\langle k^{3/2} \right\rangle }$, where ${\left\langle k \right\rangle }$ and ${\left\langle k^{3/2} \right\rangle }$ are the average degree and the $3/2$ order moment of degree distribution, respectively. Furthermore, we consider the finite size effect where the stochastic fluctuation should be involved. To the end, we derive the Langevin equation and obtain the potential of the corresponding Fokker-Planck equation. This allows us to calculate the effective critical noise $f_c(N)$ at which the susceptibility is maximal in finite size networks. We find that the $f_c-f_c(N)$ decays with $N$ in a power-law way and vanishes for $N\rightarrow \infty$. All the theoretical results are confirmed by performing the extensive Monte Carlo simulations in random $k$-regular networks, Erdös-Rényi random networks and scale-free networks.

physics.soc-ph↗

arXiv2014

Complex activated transition in a system of two coupled bistable oscillators

We study the fluctuation-activated transition process in a system of two coupled bistable oscillators, in which each oscillator is driven by one constant force and an independent Gaussian white noise. The transition pathway has been identified and the transition rate has been computed as the coupling strength $μ$ and the mismatch $σ$ in the force constants are varied. For identical oscillators ($σ=0$), the transition undergoes a change from a two-step process with two candidate pathways to a one-step process with also two candidate pathways to a one-step process with a single pathway as $μ$ is increased. For nonidentical oscillators ($σ\neq0$), a novel transition emerges that is a mixture of a two-step pathway and a one-step pathway. Interestingly, we find that the total transition rate depends nonmonotonically on $μ$: a maximal rate appears in an intermediate magnitude of $μ$. Moreover, in the presence of weak coupling the rate also exhibits an unexpected maximum as a function of $σ$. The results are in an excellent agreement with our numerical simulations by forward flux sampling.

cond-mat.stat-mech↗

arXiv2013

How does degree heterogeneity affect nucleation of Ising model on complex networks?

We investigate the nucleation of Ising model on complex networks and focus on the role played by the heterogeneity of degree distribution on nucleation rate. Using Monte Carlo simulation combined with forward flux sampling, we find that for a weak external field the nucleation rate decreases monotonically as degree heterogeneity increases. Interestingly, for a relatively strong external field the nucleation rate exhibits a nonmonotonic dependence on degree heterogeneity, in which there exists a maximal nucleation rate at an intermediate level of degree heterogeneity. Furthermore, we develop a heterogeneous mean-field theory for evaluating the free-energy barrier of nucleation. The theoretical estimations are qualitatively consistent with the simulation results. Our study suggests that degree heterogeneity plays a nontrivial role in the dynamics of phase transition in networked Ising systems.

cond-mat.stat-mech↗

arXiv2002

Statistical Error in Particle Simulations of Hydrodynamic Phenomena

We present predictions for the statistical error due to finite sampling in the presence of thermal fluctuations in molecular simulation algorithms. Specifically, we establish how these errors depend on Mach number, Knudsen number, number of particles, etc. Expressions for the common hydrodynamic variables of interest such as flow velocity, temperature, density, pressure, shear stress and heat flux are derived using equilibrium statistical mechanics. Both volume-averaged and surface-averaged quantities are considered. Comparisons between theory and computations using direct simulation Monte Carlo for dilute gases, and molecular dynamics for dense fluids, show that the use of equilibrium theory provides accurate results.

cond-mat.stat-mech↗

arXiv2001

Simulations of the Static Friction Due to Adsorbed Molecules

The static friction between crystalline surfaces separated by a molecularly thin layer of adsorbed molecules is calculated using molecular dynamics simulations. These molecules naturally lead to a finite static friction that is consistent with macroscopic friction laws. Crystalline alignment, sliding direction, and the number of adsorbed molecules are not controlled in most experiments and are shown to have little effect on the friction. Temperature, molecular geometry and interaction potentials can have larger effects on friction. The observed trends in friction can be understood in terms of a simple hard sphere model.

cond-mat.mtrl-sci↗

arXiv2000

Simulations of the kinetic friction due to adsorbed surface layers

Simulations of the kinetic friction due to a layer of adsorbed molecules between two crystalline surfaces are presented. The adsorbed layer naturally produces friction that is consistent with Amontons' laws and insensitive to parameters that are not controlled in experiments. The kinetic friction rises logarithmically with velocity as in many experimental systems. Variations with potential parameters and temperature follow variations in the static friction. This correlation is understood through analogy with the Tomlinson model and the trends are explained with a hard-sphere picture.

cond-mat.mtrl-sci↗