arXiv subjects

Guofeng Li

Publications and source records attributed to Guofeng Li.

11 recordsLinked to original sources

arXiv2024

Short-time large deviations of first-passage functionals for high-order stochastic processes

We consider high-order stochastic processes $x(t)$ described by the Langevin equation $\frac{{{d^m}x\left( t \right)}}{{d{t^m}}}= \sqrt{2D} \xi(t)$, where $\xi(t)$ is a delta-correlated Gaussian noise with zero mean, and $D$ is the strength of noise. We focus on the short-time statistics of the first-passage functionals $A=\int_{0}^{T} \left[ x(t)\right] ^n dt$ along the trajectories starting from $x(0)=L$ and terminating whenever passing through the origin for the first-time at $t=T$. Using the optimal fluctuation method, we analytically obtain the most likely realizations of the first-passage processes for a given constraint $A$ with $n=0$ and 1, corresponding to the first-passage time itself and the area swept by the first-passage trajectory, respectively. The tail of the distribution of $A$ shows an essential singularity at $A \to 0$, $P_{m,n}(A |L) \sim \exp\left(-\frac{\alpha_{m,n}L^{2mn-n+2}}{D A^{2m-1}} \right)$, where the explicit expressions for the exponents $\alpha_{m,0}$ and $\alpha_{m,1}$ for arbitrary $m$ are obtained.

cond-mat.stat-mech↗

arXiv2024

Short-time large deviation of constrained random acceleration process

By optimal fluctuation method, we study short-time distribution $P(\mathcal{A}=A)$ of the functionals, $\mathcal{A}=\int_{0}^{t_f} x^n(t) dt$, along constrained trajectories of random acceleration process for a given time duration $t_f$, where $n$ is a positive integer. We consider two types of constraints: one is called the total constraint, where the initial position and velocity and the final position and velocity are both fixed, and the other is called the partial constraint, where the initial position and velocity, the final position are fixed, and letting the final velocity be free. Via the variation of constrained action functionals, the resulting Euler-Lagrange equations are analytically solved for $n=1$ and 2, and the optimal path, i.e., the most probable realization of the random acceleration process $x(t)$, conditioned on specified $A$ and $n$, are correspondingly obtained. For $n \geq 3$, a numerical scheme is proposed to find the optimal path. We show that, for $n=1$, $P(A)$ is a Gaussian distribution with the variance proportional to $Dt_f^5$ ($D$ is the particle velocity diffusion constant). For $n \geq 2$, $P(A)$ exhibits the non-Gaussian feature. In the small-$A$ limit, $P(A)$ show a essential singularity, $-\ln P(A) \sim A^{-3}$, and the optimal path localizes around the initial state over a long-time window, and then escapes to the final position sharply at a late time. For $A$ much larger than its typical value, there are multiple optimal paths with the same $A$ but with different actions (or probability densities). Among these degenerate paths, one with the minimum action is dominant, and the others are exponentially unlikely. All the theoretical results are validated by simulating the effective Langevin equations governing the constrained random acceleration process.

cond-mat.stat-mech↗

arXiv2024

DART: Depth-Enhanced Accurate and Real-Time Background Matting

Matting with a static background, often referred to as ``Background Matting" (BGM), has garnered significant attention within the computer vision community due to its pivotal role in various practical applications like webcasting and photo editing. Nevertheless, achieving highly accurate background matting remains a formidable challenge, primarily owing to the limitations inherent in conventional RGB images. These limitations manifest in the form of susceptibility to varying lighting conditions and unforeseen shadows. In this paper, we leverage the rich depth information provided by the RGB-Depth (RGB-D) cameras to enhance background matting performance in real-time, dubbed DART. Firstly, we adapt the original RGB-based BGM algorithm to incorporate depth information. The resulting model's output undergoes refinement through Bayesian inference, incorporating a background depth prior. The posterior prediction is then translated into a "trimap," which is subsequently fed into a state-of-the-art matting algorithm to generate more precise alpha mattes. To ensure real-time matting capabilities, a critical requirement for many real-world applications, we distill the backbone of our model from a larger and more versatile BGM network. Our experiments demonstrate the superior performance of the proposed method. Moreover, thanks to the distillation operation, our method achieves a remarkable processing speed of 33 frames per second (fps) on a mid-range edge-computing device. This high efficiency underscores DART's immense potential for deployment in mobile applications}

cs.CV↗

arXiv2022

Deep Clustering: A Comprehensive Survey

Cluster analysis plays an indispensable role in machine learning and data mining. Learning a good data representation is crucial for clustering algorithms. Recently, deep clustering, which can learn clustering-friendly representations using deep neural networks, has been broadly applied in a wide range of clustering tasks. Existing surveys for deep clustering mainly focus on the single-view fields and the network architectures, ignoring the complex application scenarios of clustering. To address this issue, in this paper we provide a comprehensive survey for deep clustering in views of data sources. With different data sources and initial conditions, we systematically distinguish the clustering methods in terms of methodology, prior knowledge, and architecture. Concretely, deep clustering methods are introduced according to four categories, i.e., traditional single-view deep clustering, semi-supervised deep clustering, deep multi-view clustering, and deep transfer clustering. Finally, we discuss the open challenges and potential future opportunities in different fields of deep clustering.

cs.LG↗

arXiv2021

First passage in discrete-time absorbing Markov chains under stochastic resetting

First passage of stochastic processes under resetting has recently been an active research topic in the field of statistical physics. However, most of previous studies mainly focused on the systems with continuous time and space. In this paper, we study the effect of stochastic resetting on first passage properties of discrete-time absorbing Markov chains, described by a transition matrix $\brm{Q}$ between transient states and a transition matrix $\brm{R}$ from transient states to absorbing states. Using a renewal approach, we exactly derive the unconditional mean first passage time (MFPT) to either of absorbing states, the splitting probability the and conditional MFPT to each absorbing state. All the quantities can be expressed in terms of a deformed fundamental matrix $\brm{Z_{\gamma}}=\left[\brm{I}-(1-\gamma) \brm{Q} \right]^{-1}$ and $\brm{R}$, where $\brm{I}$ is the identity matrix, and $\gamma$ is the resetting probability at each time step. We further show a sufficient condition under which the unconditional MPFT can be optimized by stochastic resetting. Finally, we apply our results to two concrete examples: symmetric random walks on one-dimensional lattices with absorbing boundaries and voter model on complete graphs.

cond-mat.stat-mech↗

arXiv2020

Deep Embedded Multi-view Clustering with Collaborative Training

Multi-view clustering has attracted increasing attentions recently by utilizing information from multiple views. However, existing multi-view clustering methods are either with high computation and space complexities, or lack of representation capability. To address these issues, we propose deep embedded multi-view clustering with collaborative training (DEMVC) in this paper. Firstly, the embedded representations of multiple views are learned individually by deep autoencoders. Then, both consensus and complementary of multiple views are taken into account and a novel collaborative training scheme is proposed. Concretely, the feature representations and cluster assignments of all views are learned collaboratively. A new consistency strategy for cluster centers initialization is further developed to improve the multi-view clustering performance with collaborative training. Experimental results on several popular multi-view datasets show that DEMVC achieves significant improvements over state-of-the-art methods.

cs.LG↗

arXiv2019

Large deviation and anomalous fluctuations scaling in degree assortativity on configuration networks

By constructing a multicanonical Monte Carlo simulation, we obtain the full probability distribution $\rho_N(r)$ of the degree assortativity coefficient $r$ on configuration networks of size $N$ by using the multiple histogram reweighting method. We suggest that $\rho_N(r)$ obeys a large deviation principle, $\rho_N \left(r-r_N^* \right) \asymp {e^{ - {N^\xi }I\left( {r- r_N^* } \right)}}$, where the rate function $I$ is convex and possesses its unique minimum at $r=r_N^*$, and $\xi$ is an exponent that scales $\rho_N$'s with $N$. We show that $\xi=1$ for Poisson random graphs, and $\xi\geq1$ for scale-free networks in which $\xi$ is a decreasing function of the degree distribution exponent $\gamma$. Our results reveal that the fluctuations of $r$ exhibits an anomalous scaling with $N$ in highly heterogeneous networks.

cond-mat.stat-mech↗

arXiv2017

Optimal Allocation of Resources for Suppressing Epidemic Spreading on Networks

Efficient allocation of limited medical resources is crucial for controlling epidemic spreading on networks. Based on the susceptible-infected-susceptible model, we solve an optimization problem as how best to allocate the limited resources so as to minimize the prevalence, providing that the curing rate of each node is positively correlated to its medical resource. By quenched mean-field theory and heterogeneous mean-field (HMF) theory, we prove that epidemic outbreak will be suppressed to the greatest extent if the curing rate of each node is directly proportional to its degree, under which the effective infection rate $λ$ has a maximal threshold $λ_c^{opt}=1/\left\langle k \right\rangle$ where $\left\langle k \right\rangle$ is average degree of the underlying network. For weak infection region ($λ\gtrsimλ_c^{opt}$), we combine a perturbation theory with Lagrange multiplier method (LMM) to derive the analytical expression of optimal allocation of the curing rates and the corresponding minimized prevalence. For general infection region ($λ>λ_c^{opt}$), the high-dimensional optimization problem is converted into numerically solving low-dimensional nonlinear equations by the HMF theory and LMM. Counterintuitively, in the strong infection region the low-degree nodes should be allocated more medical resources than the high-degree nodes to minimize the prevalence. Finally, we use simulated annealing to validate the theoretical results.

q-bio.PE↗

arXiv2016

Epidemic extinction in a generalized susceptible-infected-susceptible model

We study the extinction of epidemics in a generalized susceptible-infected-susceptible model, where a susceptible individual becomes infected with the rate $λ$ when contacting $m$ infective individual(s) simultaneously, and an infected individual spontaneously recovers with the rate $μ$. By employing the Wentzel-Kramers-Brillouin approximation for the master equation, the problem is reduced to finding the zero-energy trajectories in an effective Hamiltonian system, and the mean extinction time $\langle T\rangle$ depends exponentially on the associated action $\mathcal {S}$ and the size of the population $N$, $\langle T\rangle \sim \exp(N\mathcal {S})$. Because of qualitatively different bifurcation features for $m=1$ and $m\geq2$, we derive independently the expressions of $\mathcal {S}$ as a function of the rescaled infection rate $λ/μ$. For the weak infection, $\mathcal {S}$ scales to the distance to the bifurcation with an exponent $2$ for $m=1$ and $3/2$ for $m\geq2$. Finally, a rare-event simulation method is used to validate the theory.

q-bio.PE↗

arXiv2016

Explosive Phase Transition in a Majority-Vote Model with Inertia

We generalize the original majority-vote model by incorporating an inertia into the microscopic dynamics of the spin flipping, where the spin-flip probability of any individual depends not only on the states of its neighbors, but also on its own state. Surprisingly, the order-disorder phase transition is changed from a usual continuous type to a discontinuous or an explosive one when the inertia is above an appropriate level. A central feature of such an explosive transition is a strong hysteresis behavior as noise intensity goes forward and backward. Within the hysteresis region, a disordered phase and two symmetric ordered phases are coexisting and transition rates between these phases are numerically calculated by a rare-event sampling method. A mean-field theory is developed to analytically reveal the property of this phase transition.

physics.soc-ph↗

arXiv2016

Discontinuous phase transition in an annealed multi-state majority-vote model

In this paper, we generalize the original majority-vote (MV) model with noise from two states to arbitrary $q$ states, where $q$ is an integer no less than two. The main emphasis is paid to the comparison on the nature of phase transitions between the two-state MV (MV2) model and the three-state MV (MV3) model. By extensive Monte Carlo simulation and mean-field analysis, we find that the MV3 model undergoes a discontinuous order-disorder phase transition, in contrast to a continuous phase transition in the MV2 model. A central feature of such a discontinuous transition is a strong hysteresis behavior as noise intensity goes forward and backward. Within the hysteresis region, the disordered phase and ordered phase are coexisting.

cond-mat.stat-mech↗