-
Asymptotic behaviors of subcritical branching killed Brownian motion with drift
Authors:
Haojie Hou,
Yan-Xia Ren,
Renming Song,
Ya** Zhu
Abstract:
In this paper, we study asymptotic behaviors of a subcritical branching killed Brownian motion with drift $-ρ$ and offspring distribution $\{p_k:k\ge 0\}$. Let $\widetildeζ^{-ρ}$ be the extinction time of this subcritical branching killed Brownian motion, $\widetilde{M}_t^{-ρ}$ the maximal position of all the particles alive at time $t$ and $\widetilde{M}^{-ρ}:=\max_{t\ge 0}\widetilde{M}_t^{-ρ}$ t…
▽ More
In this paper, we study asymptotic behaviors of a subcritical branching killed Brownian motion with drift $-ρ$ and offspring distribution $\{p_k:k\ge 0\}$. Let $\widetildeζ^{-ρ}$ be the extinction time of this subcritical branching killed Brownian motion, $\widetilde{M}_t^{-ρ}$ the maximal position of all the particles alive at time $t$ and $\widetilde{M}^{-ρ}:=\max_{t\ge 0}\widetilde{M}_t^{-ρ}$ the all time maximal position. Let $\mathbb{P}_x$ be the law of this subcritical branching killed Brownian motion when the initial particle is located at $x\in (0,\infty)$. Under the assumption $\sum_{k=1}^\infty k (\log k) p_k <\infty$, we establish the decay rates of $\mathbb{P}_x(\widetildeζ^{-ρ}>t)$ and $\mathbb{P}_x(\widetilde{M}^{-ρ}>y)$ as $t$ and $y$ tend to $\infty$ respectively. We also establish the decay rate of $\mathbb{P}_x(\widetilde{M}_t^{-ρ}>z(t,ρ))$ as $t\to\infty$, where $z(t,ρ)=\sqrt{t}z-ρt$ for $ρ\leq 0$ and $z(t,ρ)=z$ for $ρ>0$. As a consequence, we obtain a Yaglom-type limit theorem.
△ Less
Submitted 3 July, 2024; v1 submitted 1 July, 2024;
originally announced July 2024.
-
Accelerating Diffusion Models with Parallel Sampling: Inference at Sub-Linear Time Complexity
Authors:
Haoxuan Chen,
Yinuo Ren,
Lexing Ying,
Grant M. Rotskoff
Abstract:
Diffusion models have become a leading method for generative modeling of both image and scientific data. As these models are costly to train and evaluate, reducing the inference cost for diffusion models remains a major goal. Inspired by the recent empirical success in accelerating diffusion models via the parallel sampling technique~\cite{shih2024parallel}, we propose to divide the sampling proce…
▽ More
Diffusion models have become a leading method for generative modeling of both image and scientific data. As these models are costly to train and evaluate, reducing the inference cost for diffusion models remains a major goal. Inspired by the recent empirical success in accelerating diffusion models via the parallel sampling technique~\cite{shih2024parallel}, we propose to divide the sampling process into $\mathcal{O}(1)$ blocks with parallelizable Picard iterations within each block. Rigorous theoretical analysis reveals that our algorithm achieves $\widetilde{\mathcal{O}}(\mathrm{poly} \log d)$ overall time complexity, marking the first implementation with provable sub-linear complexity w.r.t. the data dimension $d$. Our analysis is based on a generalized version of Girsanov's theorem and is compatible with both the SDE and probability flow ODE implementations. Our results shed light on the potential of fast and efficient sampling of high-dimensional data on fast-evolving modern large-memory GPU clusters.
△ Less
Submitted 24 May, 2024;
originally announced May 2024.
-
Tails of extinction time and maximal displacement for critical branching killed Lévy process
Authors:
Haojie Hou,
Yan-Xia Ren,
Renming Song
Abstract:
In this paper, we study asymptotic behaviors of the tails of extinction time and maximal displacement of a critical branching killed Lévy process $(Z_t^{(0,\infty)})_{t\ge 0}$ in $\mathbb{R}$, in which all particles (and their descendants) are killed upon exiting $(0, \infty)$. Let $ζ^{(0,\infty)}$ and $M_t^{(0,\infty)}$ be the extinction time and maximal position of all the particles alive at tim…
▽ More
In this paper, we study asymptotic behaviors of the tails of extinction time and maximal displacement of a critical branching killed Lévy process $(Z_t^{(0,\infty)})_{t\ge 0}$ in $\mathbb{R}$, in which all particles (and their descendants) are killed upon exiting $(0, \infty)$. Let $ζ^{(0,\infty)}$ and $M_t^{(0,\infty)}$ be the extinction time and maximal position of all the particles alive at time $t$ of this branching killed Lévy process and define $M^{(0,\infty)}: = \sup_{t\geq 0} M_t^{(0,\infty)}$. Under the assumption that the offspring distribution belongs to the domain of attraction of an $α$-stable distribution, $α\in (1, 2]$, and some moment conditions on the spatial motion, we give the decay rates of the survival probabilities $$ \mathbb{P}_{y}(ζ^{(0,\infty)}>t), \quad \mathbb{P}_{\sqrt{t}y}(ζ^{(0,\infty)}>t) $$ and the tail probabilities $$ \mathbb{P}_{y}(M^{(0,\infty)}\geq x), \quad \mathbb{P}_{xy}(M^{(0,\infty)}\geq x). $$ We also study the scaling limits of $M_t^{(0,\infty)}$ and the point process $Z_t^{(0,\infty)}$ under $\mathbb{P}_{\sqrt{t}y}(\cdot |ζ^{(0,\infty)}>t)$ and $\mathbb{P}_y(\cdot |ζ^{(0,\infty)}>t)$. The scaling limits under $\mathbb{P}_{\sqrt{t}y}(\cdot |ζ^{(0,\infty)}>t)$ are represented in terms of super killed Brownian motion.
△ Less
Submitted 14 May, 2024;
originally announced May 2024.
-
Risk-Sensitive Mean Field Games with Common Noise: A Theoretical Study with Applications to Interbank Markets
Authors:
Xin Yue Ren,
Dena Firoozi
Abstract:
In this paper, we address linear-quadratic-Gaussian (LQG) risk-sensitive mean field games (MFGs) with common noise. In this framework agents are exposed to a common noise and aim to minimize an exponential cost functional that reflects their risk sensitivity. We leverage the convex analysis method to derive the optimal strategies of agents in the limit as the number of agents goes to infinity. The…
▽ More
In this paper, we address linear-quadratic-Gaussian (LQG) risk-sensitive mean field games (MFGs) with common noise. In this framework agents are exposed to a common noise and aim to minimize an exponential cost functional that reflects their risk sensitivity. We leverage the convex analysis method to derive the optimal strategies of agents in the limit as the number of agents goes to infinity. These strategies yield a Nash equilibrium for the limiting model. The model is then applied to interbank markets, focusing on optimizing lending and borrowing activities to assess systemic and individual bank risks when reserves drop below a critical threshold. We employ Fokker-Planck equations and the first hitting time method to formulate the overall probability of a bank or market default. We observe that the risk-averse behavior of agents reduces the probability of individual defaults and systemic risk, enhancing the resilience of the financial system. Adopting a similar approach based on stochastic Fokker-Planck equations, we further expand our analysis to investigate the conditional probabilities of individual default under specific trajectories of the common market shock.
△ Less
Submitted 6 March, 2024;
originally announced March 2024.
-
On Inhomogeneous Infinite Products of Stochastic Matrices and Applications
Authors:
Zhaoyue Xia,
Jun Du,
Chunxiao Jiang,
H. Vincent Poor,
Zhu Han,
Yong Ren
Abstract:
With the growth of magnitude of multi-agent networks, distributed optimization holds considerable significance within complex systems. Convergence, a pivotal goal in this domain, is contingent upon the analysis of infinite products of stochastic matrices (IPSMs). In this work, convergence properties of inhomogeneous IPSMs are investigated. The convergence rate of inhomogeneous IPSMs towards an abs…
▽ More
With the growth of magnitude of multi-agent networks, distributed optimization holds considerable significance within complex systems. Convergence, a pivotal goal in this domain, is contingent upon the analysis of infinite products of stochastic matrices (IPSMs). In this work, convergence properties of inhomogeneous IPSMs are investigated. The convergence rate of inhomogeneous IPSMs towards an absolute probability sequence $π$ is derived. We also show that the convergence rate is nearly exponential, which coincides with existing results on ergodic chains. The methodology employed relies on delineating the interrelations among Sarymsakov matrices, scrambling matrices, and positive-column matrices. Based on the theoretical results on inhomogeneous IPSMs, we propose a decentralized projected subgradient method for time-varying multi-agent systems with graph-related stretches in (sub)gradient descent directions. The convergence of the proposed method is established for convex objective functions, and extended to non-convex objectives that satisfy Polyak-Lojasiewicz conditions. To corroborate the theoretical findings, we conduct numerical simulations, aligning the outcomes with the established theoretical framework.
△ Less
Submitted 25 January, 2024;
originally announced January 2024.
-
QCQP-Net: Reliably Learning Feasible Alternating Current Optimal Power Flow Solutions Under Constraints
Authors:
Sihan Zeng,
Youngdae Kim,
Yuxuan Ren,
Kibaek Kim
Abstract:
At the heart of power system operations, alternating current optimal power flow (ACOPF) studies the generation of electric power in the most economical way under network-wide load requirement, and can be formulated as a highly structured non-convex quadratically constrained quadratic program (QCQP). Optimization-based solutions to ACOPF (such as ADMM or interior-point method), as the classic appro…
▽ More
At the heart of power system operations, alternating current optimal power flow (ACOPF) studies the generation of electric power in the most economical way under network-wide load requirement, and can be formulated as a highly structured non-convex quadratically constrained quadratic program (QCQP). Optimization-based solutions to ACOPF (such as ADMM or interior-point method), as the classic approach, require large amount of computation and cannot meet the need to repeatedly solve the problem as load requirement frequently changes. On the other hand, learning-based methods that directly predict the ACOPF solution given the load input incur little computational cost but often generates infeasible solutions (i.e. violate the constraints of ACOPF). In this work, we combine the best of both worlds -- we propose an innovated framework for learning ACOPF, where the input load is mapped to the ACOPF solution through a neural network in a computationally efficient and reliable manner. Key to our innovation is a specific-purpose "activation function" defined implicitly by a QCQP and a novel loss, which enforce constraint satisfaction. We show through numerical simulations that our proposed method achieves superior feasibility rate and generation cost in situations where the existing learning-based approaches fail.
△ Less
Submitted 11 January, 2024;
originally announced January 2024.
-
From 0 to 3: Intermediate phases between normal and anomalous spreading of two-type branching Brownian motion
Authors:
Heng Ma,
Yan-Xia Ren
Abstract:
The logarithmic correction for the order of the maximum of a two-type reducible branching Brownian motion on the real line exhibits a double jump when the parameters (the ratio of the diffusion coefficients of the two types of particles, and the ratio of the branching rates the two types of particles) cross the boundary of the anomalous spreading region identified by Biggins.
In this paper, we f…
▽ More
The logarithmic correction for the order of the maximum of a two-type reducible branching Brownian motion on the real line exhibits a double jump when the parameters (the ratio of the diffusion coefficients of the two types of particles, and the ratio of the branching rates the two types of particles) cross the boundary of the anomalous spreading region identified by Biggins.
In this paper, we further examine this double jump phenomenon by studying a two-type reducible branching Brownian motion on the real line with its parameters depend on the time horizon t. We show that when the parameters approach the boundaries of the anomalous spreading region in an appropriate way, the order of the maximum can interpolate smoothly between different surrounding regimes. We also determine the asymptotic law of the maximum and characterize the extremal process.
△ Less
Submitted 21 December, 2023;
originally announced December 2023.
-
Statistical Spatially Inhomogeneous Diffusion Inference
Authors:
Yinuo Ren,
Yi** Lu,
Lexing Ying,
Grant M. Rotskoff
Abstract:
Inferring a diffusion equation from discretely-observed measurements is a statistical challenge of significant importance in a variety of fields, from single-molecule tracking in biophysical systems to modeling financial instruments. Assuming that the underlying dynamical process obeys a $d$-dimensional stochastic differential equation of the form…
▽ More
Inferring a diffusion equation from discretely-observed measurements is a statistical challenge of significant importance in a variety of fields, from single-molecule tracking in biophysical systems to modeling financial instruments. Assuming that the underlying dynamical process obeys a $d$-dimensional stochastic differential equation of the form $$\mathrm{d}\boldsymbol{x}_t=\boldsymbol{b}(\boldsymbol{x}_t)\mathrm{d} t+Σ(\boldsymbol{x}_t)\mathrm{d}\boldsymbol{w}_t,$$ we propose neural network-based estimators of both the drift $\boldsymbol{b}$ and the spatially-inhomogeneous diffusion tensor $D = ΣΣ^{T}$ and provide statistical convergence guarantees when $\boldsymbol{b}$ and $D$ are $s$-Hölder continuous. Notably, our bound aligns with the minimax optimal rate $N^{-\frac{2s}{2s+d}}$ for nonparametric function estimation even in the presence of correlation within observational data, which necessitates careful handling when establishing fast-rate generalization bounds. Our theoretical results are bolstered by numerical experiments demonstrating accurate inference of spatially-inhomogeneous diffusion tensors.
△ Less
Submitted 10 December, 2023;
originally announced December 2023.
-
Generic root counts of tropically transverse systems -- An invitation to tropical geometry in OSCAR
Authors:
Isaac Holt,
Yue Ren
Abstract:
The main mathematical focus of this paper is a class of parametrised polynomial systems that we refer to as being tropically transverse. We show how their generic number of solutions can be expressed as the mixed volume of a modified system. We then provide an alternate proof of a recent result by Borovik et al, on the number of equilibria of coupled nonlinear oscillators using elementary tropical…
▽ More
The main mathematical focus of this paper is a class of parametrised polynomial systems that we refer to as being tropically transverse. We show how their generic number of solutions can be expressed as the mixed volume of a modified system. We then provide an alternate proof of a recent result by Borovik et al, on the number of equilibria of coupled nonlinear oscillators using elementary tropical geometry. The proof draws upon a wide range of concepts across tropical geometry, which we will use as an opportunity to give a first overview over various tropical features in OSCAR.
△ Less
Submitted 29 November, 2023;
originally announced November 2023.
-
$1$-stable fluctuation of the derivative martingale of branching random walk
Authors:
Haojie Hou,
Yan-Xia Ren,
Renming Song
Abstract:
In this paper, we study the functional convergence in law of the fluctuations of the derivative martingale of branching random walk on the real line. Our main result strengthens the results of Buraczewski et. al. [Ann. Probab., 2021] and is the branching random walk counterpart of the main result of Maillard and Pain [Ann. Probab., 2019] for branching Brownian motion.
In this paper, we study the functional convergence in law of the fluctuations of the derivative martingale of branching random walk on the real line. Our main result strengthens the results of Buraczewski et. al. [Ann. Probab., 2021] and is the branching random walk counterpart of the main result of Maillard and Pain [Ann. Probab., 2019] for branching Brownian motion.
△ Less
Submitted 27 November, 2023;
originally announced November 2023.
-
Multi-Objective Optimization via Wasserstein-Fisher-Rao Gradient Flow
Authors:
Yinuo Ren,
Tesi Xiao,
Tanmay Gangwani,
Anshuka Rangi,
Holakou Rahmanian,
Lexing Ying,
Subhajit Sanyal
Abstract:
Multi-objective optimization (MOO) aims to optimize multiple, possibly conflicting objectives with widespread applications. We introduce a novel interacting particle method for MOO inspired by molecular dynamics simulations. Our approach combines overdamped Langevin and birth-death dynamics, incorporating a "dominance potential" to steer particles toward global Pareto optimality. In contrast to pr…
▽ More
Multi-objective optimization (MOO) aims to optimize multiple, possibly conflicting objectives with widespread applications. We introduce a novel interacting particle method for MOO inspired by molecular dynamics simulations. Our approach combines overdamped Langevin and birth-death dynamics, incorporating a "dominance potential" to steer particles toward global Pareto optimality. In contrast to previous methods, our method is able to relocate dominated particles, making it particularly adept at managing Pareto fronts of complicated geometries. Our method is also theoretically grounded as a Wasserstein-Fisher-Rao gradient flow with convergence guarantees. Extensive experiments confirm that our approach outperforms state-of-the-art methods on challenging synthetic and real-world datasets.
△ Less
Submitted 21 November, 2023;
originally announced November 2023.
-
Asymptotic expansion for additive measure of branching Brownian motion
Authors:
Haojie Hou,
Yan-Xia Ren,
Renming Song
Abstract:
Let $N(t)$ be the collection of particles alive at time $t$ in a branching Brownian motion in $\mathbb{R}^d$, and for $u\in N(t)$, let $\mathbf{X}_u(t)$ be the position of particle $u$ at time $t$. For $θ\in \mathbb{R}^d$, we define the additive measures of the branching Brownian motion by…
▽ More
Let $N(t)$ be the collection of particles alive at time $t$ in a branching Brownian motion in $\mathbb{R}^d$, and for $u\in N(t)$, let $\mathbf{X}_u(t)$ be the position of particle $u$ at time $t$. For $θ\in \mathbb{R}^d$, we define the additive measures of the branching Brownian motion by$$μ_t^θ(\mathrm{d}\mathbf{x}):= e^{-(1+\frac{\Vertθ\Vert^2}{2})t}\sum_{u\in N(t)} e^{-θ\cdot \mathbf{X}_u(t)} δ_{\left(\mathbf{X}_u(t)+θt\right)}(\mathrm{d}\mathbf{x}).$$ In this paper, under some conditions on the offspring distribution, we give asymptotic expansions of arbitrary order for $μ_t^θ((\mathbf{a}, \mathbf{b}])$ and $μ_t^θ((-\infty, \mathbf{a}])$ for $θ\in \mathbb{R}^d$ with $\Vert θ\Vert <\sqrt{2}$. These expansions sharpen the asymptotic results of Asmussen and Kaplan (1976) and Kang (1999), and are analogs of the expansions in Gao and Liu (2021) and Révész, Rosen and Shi (2005) for branching Wiener processes (a particular class of branching random walks) corresponding to $θ=\mathbf{0}$.
△ Less
Submitted 28 October, 2023;
originally announced October 2023.
-
Tail probability of maximal displacement in critical branching Lévy process with stable branching
Authors:
Haojie Hou,
Yiyang Jiang,
Yan-Xia Ren,
Renming Song
Abstract:
Consider a critical branching Lévy process $\{X_t, t\ge 0\}$ with branching rate $β>0, $ offspring distribution $\{p_k:k\geq 0\}$ and spatial motion $\{ξ_t, Π_x\}$. For any $t\ge 0$, let $N_t$ be the collection of particles alive at time $t$, and, for any $u\in N_t$, let $X_u(t)$ be the position of $u$ at time $t$. We study the tail probability of the maximal displacement…
▽ More
Consider a critical branching Lévy process $\{X_t, t\ge 0\}$ with branching rate $β>0, $ offspring distribution $\{p_k:k\geq 0\}$ and spatial motion $\{ξ_t, Π_x\}$. For any $t\ge 0$, let $N_t$ be the collection of particles alive at time $t$, and, for any $u\in N_t$, let $X_u(t)$ be the position of $u$ at time $t$. We study the tail probability of the maximal displacement $M:=\sup_{t>0}\sup_{u\in N_t} X_u(t)$ under the assumption $\lim_{n\to\infty} n^α\sum_{k=n}^\infty p_k =κ\in(0,\infty)$ for some $α\in (1,2)$, $Π_0(ξ_1)=0$ and $Π_0 (|ξ_1|^r)\in (0,\infty)$ for some $r> 2α/(α-1)$. Our main result is a generalization of the main result of Sawyer and Fleischman (1979) for branching Brownian motions and that of Lalley and Shao (2015) for branching random walks, both of which are proved under the assumption $\sum_{k=0}^\infty k^3 p_k<\infty$.
△ Less
Submitted 8 October, 2023;
originally announced October 2023.
-
Stationary measures and the continuous-state branching process conditioned on extinction
Authors:
Rongli Liu,
Yan-Xia Ren,
Ting Yang
Abstract:
We consider continuous-state branching processes (CB processes) which become extinct almost surely. First, we tackle the problem of describing the stationary measures on $(0,+\infty)$ for such CB processes. We give a representation of the stationary measure in terms of scale functions of related Lévy processes. Then we prove that the stationary measure can be obtained from the vague limit of the p…
▽ More
We consider continuous-state branching processes (CB processes) which become extinct almost surely. First, we tackle the problem of describing the stationary measures on $(0,+\infty)$ for such CB processes. We give a representation of the stationary measure in terms of scale functions of related Lévy processes. Then we prove that the stationary measure can be obtained from the vague limit of the potential measure, and, in the critical case, can also be obtained from the vague limit of a normalized transition probability. Next, we prove some limit theorems for the CB process conditioned on extinction in a near future and on extinction at a fixed time. We obtain non-degenerate limit distributions which are of the size-biased type of the stationary measure in the critical case and of the Yaglom's distribution in the subcritical case. Finally we explore some further properties of the limit distributions.
△ Less
Submitted 3 September, 2023;
originally announced September 2023.
-
Asymptotic expansion for branching killed Brownian motion with drift
Authors:
Haojie Hou,
Yan-Xia Ren,
Renming Song
Abstract:
Let $Z_t^{(0,\infty)}$ be the point process formed by the positions of all particles alive at time $t$ in a branching Brownian motion with drift and killed upon reaching 0. We study the asymptotic expansions of $Z_t^{(0,\infty)}(A)$ for $A= (a,b)$ and $A=(a,\infty)$ under the assumption that $\sum_{k=1}^\infty k(\log k)^{1+λ} p_k <\infty$ for large $λ$ in the regime of $θ\in [0,\sqrt{2})$. These r…
▽ More
Let $Z_t^{(0,\infty)}$ be the point process formed by the positions of all particles alive at time $t$ in a branching Brownian motion with drift and killed upon reaching 0. We study the asymptotic expansions of $Z_t^{(0,\infty)}(A)$ for $A= (a,b)$ and $A=(a,\infty)$ under the assumption that $\sum_{k=1}^\infty k(\log k)^{1+λ} p_k <\infty$ for large $λ$ in the regime of $θ\in [0,\sqrt{2})$. These results extend and sharpen the results of Louidor and Saglietti [J. Stat. Phys, 2020] and Kesten [Stochastic Process. Appl., 1978].
△ Less
Submitted 20 July, 2023;
originally announced July 2023.
-
Coalescence times for critical Galton-Watson processes with immigration
Authors:
Rong-Li Liu,
Yan-Xia Ren,
Yingrui Wang
Abstract:
Let $X^I_n$ be the coalescence time of two particles picked at random from the $n$th generation of a critical Galton-Watson process with immigration, and let $A^I_n$ be the coalescence time of the whole population in the $n$th generation. In this paper, we study the limiting behaviors of $X^I_n$ and $A^I_n$ as $n\to\infty$.
Let $X^I_n$ be the coalescence time of two particles picked at random from the $n$th generation of a critical Galton-Watson process with immigration, and let $A^I_n$ be the coalescence time of the whole population in the $n$th generation. In this paper, we study the limiting behaviors of $X^I_n$ and $A^I_n$ as $n\to\infty$.
△ Less
Submitted 14 July, 2023;
originally announced July 2023.
-
Modular representations of strange classical Lie superalgebras and the first super Kac-Weisfeiler conjecture
Authors:
Ye Ren,
Bin Shu,
Fanlei Yang,
An Zhang
Abstract:
Suppose $\mathfrak{g}=\mathfrak{g}_{\bar 0}+\mathfrak{g}_{\bar 1} is a Lie superalgebra of queer type or periplectic type over an algebraically closed field $\textbf{k}$ of characteristic $p>2…
▽ More
Suppose $\mathfrak{g}=\mathfrak{g}_{\bar 0}+\mathfrak{g}_{\bar 1} is a Lie superalgebra of queer type or periplectic type over an algebraically closed field $\textbf{k}$ of characteristic $p>2$. In this article, we initiate preliminarily to investigate modular representations of periplectic Lie superalgebras
and then verify the first super Kac-Weisfeiler conjecture on the maximal dimensions of irreducible modules for $\mathfrak{g}$ proposed by the second-named author in [Shu] where the conjecture is targeted at all finite-dimensional restricted Lie superalgebras over $\bk$, and already proved to be true for basic classical Lie superalgebras and completely solvable restricted Lie superalgebras.
△ Less
Submitted 2 July, 2023;
originally announced July 2023.
-
Limiting distributions for a class of super-Brownian motions with spatially dependent branching mechanisms
Authors:
Yan-Xia Ren,
Ting Yang
Abstract:
In this paper we consider a large class of super-Brownian motions in $\mathbb{R}$ with spatially dependent branching mechanisms. We establish the almost sure growth rate of the mass located outside a time-dependent interval $(-δt,δt)$ for $δ>0$. The growth rate is given in terms of the principal eigenvalue $λ_{1}$ of the Schödinger type operator associated with the branching mechanism. From this r…
▽ More
In this paper we consider a large class of super-Brownian motions in $\mathbb{R}$ with spatially dependent branching mechanisms. We establish the almost sure growth rate of the mass located outside a time-dependent interval $(-δt,δt)$ for $δ>0$. The growth rate is given in terms of the principal eigenvalue $λ_{1}$ of the Schödinger type operator associated with the branching mechanism. From this result we see the existence of phase transition for the growth order at $δ=\sqrt{λ_{1}/2}$. We further show that the super-Brownian motion shifted by $\sqrt{λ_{1}/2}\,t$ converges in distribution to a random measure with random density mixed by a martingale limit.
△ Less
Submitted 14 June, 2023;
originally announced June 2023.
-
Double jump in the maximum of two-type reducible branching Brownian motion
Authors:
Heng Ma,
Yan-Xia Ren
Abstract:
Consider a two-type reducible branching Brownian motion in which particles' diffusion coefficients and branching rates are influenced by their types. Here reducible means that type 1 particles can produce particles of type 1 and type 2, but type 2 particles can only produce particles of type 2. The maximum of this process is determined by two parameters: the ratio of the diffusion coefficients and…
▽ More
Consider a two-type reducible branching Brownian motion in which particles' diffusion coefficients and branching rates are influenced by their types. Here reducible means that type 1 particles can produce particles of type 1 and type 2, but type 2 particles can only produce particles of type 2. The maximum of this process is determined by two parameters: the ratio of the diffusion coefficients and the ratio of the branching rates for particles of different types. Belloum and Mallein [Electron. J. Probab. 26(20121), no. 61] identified three phases of the maximum and the extremal process, corresponding to three regions in the parameter space.
We investigate how the extremal process behaves asymptotically when the parameters lie on the boundaries between these regions. An interesting consequence is that a double jump occurs in the maximum when the parameters cross the boundary of the so called anomalous spreading region, while only single jump occurs when the parameters cross the boundary between the remaining two regions.
△ Less
Submitted 26 May, 2023; v1 submitted 17 May, 2023;
originally announced May 2023.
-
Hands-on Tropical Geometry
Authors:
Herbert Gangl,
Yue Ren,
Ziva Urbancic
Abstract:
This is a guide on how to create 3d printable models of tropical surfaces, curves, and combinations thereof. It uses Polymake to construct bounded models of the tropical objects, and OpenSCAD to thicken and export them to any common 3D printable file format.
This is a guide on how to create 3d printable models of tropical surfaces, curves, and combinations thereof. It uses Polymake to construct bounded models of the tropical objects, and OpenSCAD to thicken and export them to any common 3D printable file format.
△ Less
Submitted 20 April, 2023;
originally announced April 2023.
-
Depth Separation with Multilayer Mean-Field Networks
Authors:
Yunwei Ren,
Mo Zhou,
Rong Ge
Abstract:
Depth separation -- why a deeper network is more powerful than a shallower one -- has been a major problem in deep learning theory. Previous results often focus on representation power. For example, arXiv:1904.06984 constructed a function that is easy to approximate using a 3-layer network but not approximable by any 2-layer network. In this paper, we show that this separation is in fact algorithm…
▽ More
Depth separation -- why a deeper network is more powerful than a shallower one -- has been a major problem in deep learning theory. Previous results often focus on representation power. For example, arXiv:1904.06984 constructed a function that is easy to approximate using a 3-layer network but not approximable by any 2-layer network. In this paper, we show that this separation is in fact algorithmic: one can learn the function constructed by arXiv:1904.06984 using an overparameterized network with polynomially many neurons efficiently. Our result relies on a new way of extending the mean-field limit to multilayer networks, and a decomposition of loss that factors out the error introduced by the discretization of infinite-width mean-field networks.
△ Less
Submitted 3 April, 2023;
originally announced April 2023.
-
Extremal process for irreducible multitype branching Brownian motion
Authors:
Haojie Hou,
Yan-Xia Ren,
Renming Song
Abstract:
We first study the convergence of solutions of a system of F-KPP equations related to irreducible multitype branching Brownian motions with Heaviside-type initial conditions to traveling wave solutions. Then we apply this convergence result to prove that the extremal processes of irreducible multitype branching Brownian motions converges weakly to a cluster point process.
We first study the convergence of solutions of a system of F-KPP equations related to irreducible multitype branching Brownian motions with Heaviside-type initial conditions to traveling wave solutions. Then we apply this convergence result to prove that the extremal processes of irreducible multitype branching Brownian motions converges weakly to a cluster point process.
△ Less
Submitted 22 March, 2023; v1 submitted 21 March, 2023;
originally announced March 2023.
-
Existence and regularity results for anisotropic parabolic equations with degenerate coercivity
Authors:
Weilin Zou,
Yuanchun Ren,
Wei Wang
Abstract:
This paper deals with a class of nonlinear anisotropic parabolic equations with degenerate coercivity. Using the anisotropic Gagliardo-Nirenberg-type inequality, we prove some existence and regularity results for the solutions under the framework of anisotropic Sobolev spaces, which generalize the previous results of [10,18,23].
This paper deals with a class of nonlinear anisotropic parabolic equations with degenerate coercivity. Using the anisotropic Gagliardo-Nirenberg-type inequality, we prove some existence and regularity results for the solutions under the framework of anisotropic Sobolev spaces, which generalize the previous results of [10,18,23].
△ Less
Submitted 16 March, 2023;
originally announced March 2023.
-
On intersections and stable intersections of tropical hypersurfaces
Authors:
Yue Ren
Abstract:
We prove that every connected component of an intersection of tropical hypersurfaces contains a point of their stable intersection unless their stable intersection is empty. This is done by studying algebraic hypersurfaces that tropicalize to them and the tropicalization of their intersection.
We prove that every connected component of an intersection of tropical hypersurfaces contains a point of their stable intersection unless their stable intersection is empty. This is done by studying algebraic hypersurfaces that tropicalize to them and the tropicalization of their intersection.
△ Less
Submitted 23 February, 2023;
originally announced February 2023.
-
Gradient Estimates For The CR Heat Equation On Complete noncompact Pseudo-Hermitian Manifolds
Authors:
Yuxin Dong,
Yibin Ren,
Biqiang Zhao
Abstract:
In this paper, we derive local and global Li-Yau type gradient estimates for the positive solutions of the CR heat equation on complete noncompact pseudo-Hermitian manifolds. As applications of the gradient estimates, we give a Harnack inequality for the positive solutions of the CR heat equation, and then obtain an upper bound estimate for the corresponding heat kernel.
In this paper, we derive local and global Li-Yau type gradient estimates for the positive solutions of the CR heat equation on complete noncompact pseudo-Hermitian manifolds. As applications of the gradient estimates, we give a Harnack inequality for the positive solutions of the CR heat equation, and then obtain an upper bound estimate for the corresponding heat kernel.
△ Less
Submitted 9 May, 2023; v1 submitted 19 February, 2023;
originally announced February 2023.
-
Spectral conditions for forbidden subgraphs in bipartite graphs
Authors:
Yuan Ren,
**g Zhang,
Zhiyuan Zhang
Abstract:
A graph $G$ is $H$-free, if it contains no $H$ as a subgraph. A graph is said to be \emph{$H$-minor free}, if it does not contain $H$ as a minor. In recent years, Nikiforov asked that what is the maximum spectral radius of an $H$-free graph of order $n$? In this paper, we consider about some Brualdi-Solheid-Turán type problems on bipartite graphs. In 2015, Zhai, Lin and Gong proved that if $G$ is…
▽ More
A graph $G$ is $H$-free, if it contains no $H$ as a subgraph. A graph is said to be \emph{$H$-minor free}, if it does not contain $H$ as a minor. In recent years, Nikiforov asked that what is the maximum spectral radius of an $H$-free graph of order $n$? In this paper, we consider about some Brualdi-Solheid-Turán type problems on bipartite graphs. In 2015, Zhai, Lin and Gong proved that if $G$ is a bipartite graph with order $n \geq 2k+2$ and $ρ(G)\geq ρ(K_{k,n-k})$, then $G$ contains a $C_{2k+2}$ unless $G \cong K_{k,n-k}$ [Linear Algebra Appl. 471 (2015)]. Firstly, we give a new and more simple proof for the above theorem. Secondly, we prove that if $G$ is a bipartite graph with order $n \geq 2k+2$ and $ρ(G)\geq ρ(K_{k,n-k})$, then $G$ contains all $T_{2k+3}$ unless $G \cong K_{k,n-k}$. Finally, we prove that among all outerplanar bipartite graphs on $n>344569$ vertices, $K_{1,n-1}$ attains the maximum spectral radius.
△ Less
Submitted 7 February, 2023;
originally announced February 2023.
-
Lower deviation for the supremum of the support of super-Brownian motion
Authors:
Yan-Xia Ren,
Renming Song,
Rui Zhang
Abstract:
We study the asymptotic behavior of the supremum $M_t$ of the support of a supercritical super-Brownian motion. In our recent paper (Stoch. Proc. Appl. 137 (2021), 1-34), we showed that, under some conditions, $M_t-m(t)$ converges in distribution to a randomly shifted Gumbel random variable, where $m(t)=c_0t-c_1\log t$. In the same paper, we also studied the upper large deviation of $M_t$, i.e., t…
▽ More
We study the asymptotic behavior of the supremum $M_t$ of the support of a supercritical super-Brownian motion. In our recent paper (Stoch. Proc. Appl. 137 (2021), 1-34), we showed that, under some conditions, $M_t-m(t)$ converges in distribution to a randomly shifted Gumbel random variable, where $m(t)=c_0t-c_1\log t$. In the same paper, we also studied the upper large deviation of $M_t$, i.e., the asymptotic behavior of $\mathbb{P}(M_t>δc_0t) $ for $δ\ge 1$. In this paper, we study the lower large deviation of $M_t$, i.e., the asymptotic behavior of $\mathbb{P}(M_t\le δc_0t|\mathcal{S}) $ for $δ<1$, where $\mathcal{S}$ is the survival event.
△ Less
Submitted 8 December, 2022;
originally announced December 2022.
-
High-dimensional density estimation with tensorizing flow
Authors:
Yinuo Ren,
Hongli Zhao,
Yuehaw Khoo,
Lexing Ying
Abstract:
We propose the tensorizing flow method for estimating high-dimensional probability density functions from the observed data. The method is based on tensor-train and flow-based generative modeling. Our method first efficiently constructs an approximate density in the tensor-train form via solving the tensor cores from a linear system based on the kernel density estimators of low-dimensional margina…
▽ More
We propose the tensorizing flow method for estimating high-dimensional probability density functions from the observed data. The method is based on tensor-train and flow-based generative modeling. Our method first efficiently constructs an approximate density in the tensor-train form via solving the tensor cores from a linear system based on the kernel density estimators of low-dimensional marginals. We then train a continuous-time flow model from this tensor-train density to the observed empirical distribution by performing a maximum likelihood estimation. The proposed method combines the optimization-less feature of the tensor-train with the flexibility of the flow-based generative models. Numerical results are included to demonstrate the performance of the proposed method.
△ Less
Submitted 1 December, 2022;
originally announced December 2022.
-
Central limit theorem and Berry-Esseen bounds for a branching random walk with immigration in a random environment
Authors:
Chunmao Huang,
Yukun Ren,
Runze Li
Abstract:
We consider a branching random walk on $d$-dimensional real space with immigration in a time-dependent random environment. Let $Z_n(\mathbf t)$ be the so-called partition function of the process, namely, the moment generating function of the counting measure describing the dispersion of individuals at time $n$. For $\mathbf t$ fixed, the logarithm $\log Z_n(\mathbf t)$ satisfies a central limit th…
▽ More
We consider a branching random walk on $d$-dimensional real space with immigration in a time-dependent random environment. Let $Z_n(\mathbf t)$ be the so-called partition function of the process, namely, the moment generating function of the counting measure describing the dispersion of individuals at time $n$. For $\mathbf t$ fixed, the logarithm $\log Z_n(\mathbf t)$ satisfies a central limit theorem. By studying the logarithmic moments of the intrinsic submartingale of the system and its convergence rates, we establish the uniform and non-uniform Berry-Esseen bounds corresponding to the central limit theorem, and discover the exact convergence rate in the central limit theorem.
△ Less
Submitted 14 October, 2022;
originally announced October 2022.
-
Weak convergence of the extremes of branching Lévy processes with regularly varying tails
Authors:
Yan-Xia Ren,
Renming Song,
Rui Zhang
Abstract:
In this paper, we study the weak convergence of the extremes of supercritical branching Lévy processes $\{\mathbb{X}_t, t \ge0\}$ whose spatial motions are Lévy processes with regularly varying tails. The result is drastically different from the case of branching Brownian motions. We prove that, when properly renormalized, $\mathbb{X}_t$ converges weakly. As a consequence, we obtain a limit theore…
▽ More
In this paper, we study the weak convergence of the extremes of supercritical branching Lévy processes $\{\mathbb{X}_t, t \ge0\}$ whose spatial motions are Lévy processes with regularly varying tails. The result is drastically different from the case of branching Brownian motions. We prove that, when properly renormalized, $\mathbb{X}_t$ converges weakly. As a consequence, we obtain a limit theorem for the order statistics of $\mathbb{X}_t$.
△ Less
Submitted 12 October, 2022;
originally announced October 2022.
-
The extremal process of super-Brownian motion: a probabilistic approach via skeletons
Authors:
Yan-Xia Ren,
Ting Yang,
Rui Zhang
Abstract:
Recently Ren et al. [Stoch. Proc. Appl., 137 (2021)] have proved that the extremal process of the super-Brownian motion converges in distribution in the limit of large times. Their techniques rely heavily on the study of the convergence of solutions to the Kolmogorov-Petrovsky-Piscounov equation along the lines of [M. Bramson, Mem. Amer. Math. Soc., 44 (1983)]. In this paper we take a different ap…
▽ More
Recently Ren et al. [Stoch. Proc. Appl., 137 (2021)] have proved that the extremal process of the super-Brownian motion converges in distribution in the limit of large times. Their techniques rely heavily on the study of the convergence of solutions to the Kolmogorov-Petrovsky-Piscounov equation along the lines of [M. Bramson, Mem. Amer. Math. Soc., 44 (1983)]. In this paper we take a different approach. Our approach is based on the skeleton decomposition of super-Brownian motion. The skeleton may be interpreted as immortal particles that determine the large time behaviour of the process. We exploit this fact and carry asymptotic properties from the skeleton over to the super-Brownian motion. Some new results concerning the probabilistic representations of the limiting process are obtained, which cannot be directly obtained through the results of [Y.-X. Ren et al., Stoch. Proc. Appl., 137 (2021)]. Apart from the results, our approach offers insights into the driving force behind the limiting process for super-Brownian motions.
△ Less
Submitted 31 August, 2022;
originally announced August 2022.
-
Invariance principle for the maximal position process of branching Brownian motion in random environment
Authors:
Haojie Hou,
Yan-Xia Ren,
Renming Song
Abstract:
In this paper we study the maximal position process of branching Brownian motion in random spatial environment. The random environment is given by a process $ξ= \left(ξ(x)\right)_{x\in\mathbb{R}}$ satisfying certain conditions. We show that the maximum position $M_t$ of particles alive at time $t$ satisfies a quenched strong law of large numbers and an annealed invariance principle.
In this paper we study the maximal position process of branching Brownian motion in random spatial environment. The random environment is given by a process $ξ= \left(ξ(x)\right)_{x\in\mathbb{R}}$ satisfying certain conditions. We show that the maximum position $M_t$ of particles alive at time $t$ satisfies a quenched strong law of large numbers and an annealed invariance principle.
△ Less
Submitted 16 June, 2022;
originally announced June 2022.
-
Generic root counts and flatness in tropical geometry
Authors:
Paul Alexander Helminck,
Yue Ren
Abstract:
We use tropical and non-archimedean geometry to study generic root counts of families of polynomial equations. These families are given as morphisms of schemes $X\to{Y}$ that factor through a closed embedding into a relative torus over a parameter space $Y$. We prove a generalization of Bernstein's theorem for these morphisms, showing that the root count of a single well-behaved tropical fiber spr…
▽ More
We use tropical and non-archimedean geometry to study generic root counts of families of polynomial equations. These families are given as morphisms of schemes $X\to{Y}$ that factor through a closed embedding into a relative torus over a parameter space $Y$. We prove a generalization of Bernstein's theorem for these morphisms, showing that the root count of a single well-behaved tropical fiber spreads to an open dense subset of $Y$. We use this to express the generic root count of a wide class of square systems in terms of the matroidal degree of an explicit variety. This in particular gives tropical formulas for the birational intersection indices and volumes of Newton-Okounkov bodies defined by Kaveh and Khovanskii, and the generic root counts of the steady-state equations of chemical reaction networks. An important role in these theorems is played by the notion of tropical flatness, which allows us to infer generic properties of $X\to{Y}$ from a single tropical fiber. We show that the tropical analogue of the generic flatness theorem holds, in the sense that $X\to{Y}$ is tropically flat over an open dense subset of the Berkovich analytification of $Y$.
△ Less
Submitted 14 October, 2022; v1 submitted 15 June, 2022;
originally announced June 2022.
-
A new extension of generalized Drazin inverse in Banach algebras
Authors:
Yanxun Ren,
Lining Jiang
Abstract:
In this paper, we introduce and study a new generalized inverse, called ag-Drazin inverses in a Banach algebra $\mathcal{A}$ with unit $1$. An element $a\in\mathcal{A}$ is ag-Drazin invertible if there exists $x\in\mathcal{A}$ such that $ax=xa, \, xax=x \ {\rm and} \ a-axa\in\mathcal{A}^{acc}$, where…
▽ More
In this paper, we introduce and study a new generalized inverse, called ag-Drazin inverses in a Banach algebra $\mathcal{A}$ with unit $1$. An element $a\in\mathcal{A}$ is ag-Drazin invertible if there exists $x\in\mathcal{A}$ such that $ax=xa, \, xax=x \ {\rm and} \ a-axa\in\mathcal{A}^{acc}$, where $\mathcal{A}^{acc}\triangleq\{a\in\mathcal{A}: a-λ1 \ {\rm is \ generalized \ Drazin\ invertible} \ {\rm for \ all} \ λ\in\mathbb{C}\backslash\{0\}\}.$ Using idempotent elements, we characterize this inverse and give some its representations. Also, we prove that $a\in\mathcal{A}$ is ag-Drazin invertible if and only if $0$ is not an accumulation point of $σ_{d}(a)$, where $σ_{d}(a)$ is the generalized Drazin spectrum of $a$.
△ Less
Submitted 7 May, 2022;
originally announced May 2022.
-
A Pressure Correction Projection Finite Element Method for The 2D/3D Time-Dependent Thermomicropolar Fluid Problem
Authors:
Yuhang Ren,
Demin Liu
Abstract:
In this paper, the pressure correctionfinite element method is proposed for the 2D/3D time-dependent thermomicropolarfluid equations. Thefirst-order and second-order backward difference formulas (BDF) are adopted to approximate the time derivative term, stability analysis and error estimation of the first-order semi-discrete scheme are proved. Finally, some numerical examples are given to show the…
▽ More
In this paper, the pressure correctionfinite element method is proposed for the 2D/3D time-dependent thermomicropolarfluid equations. Thefirst-order and second-order backward difference formulas (BDF) are adopted to approximate the time derivative term, stability analysis and error estimation of the first-order semi-discrete scheme are proved. Finally, some numerical examples are given to show the effectiveness and reliability of the proposed method, which can be used to simulate the problem with high Rayleigh number.
△ Less
Submitted 29 March, 2022;
originally announced March 2022.
-
Triangular Prism Equations and Categorification
Authors:
Zhengwei Liu,
Sebastien Palcoux,
Yunxiang Ren
Abstract:
We introduce the triangular prism equations for fusion categories, which turn out to be equivalent to the pentagon equations in the spherical case (up to a change of basis), but provide insight to manage the complexity by localization. We also prove a conjecture of Z. Wang about the second Frobenius-Schur indicator on pivotal fusion categories. As applications, we provide new categorification crit…
▽ More
We introduce the triangular prism equations for fusion categories, which turn out to be equivalent to the pentagon equations in the spherical case (up to a change of basis), but provide insight to manage the complexity by localization. We also prove a conjecture of Z. Wang about the second Frobenius-Schur indicator on pivotal fusion categories. As applications, we provide new categorification criteria and complete the classification of the unitary simple integral fusion categories of Frobenius type up to rank 8 and FPdim 4080.
△ Less
Submitted 29 May, 2023; v1 submitted 12 March, 2022;
originally announced March 2022.
-
$\ell_1$DecNet+: A new architecture framework by $\ell_1$ decomposition and iteration unfolding for sparse feature segmentation
Authors:
Yumeng Ren,
Yiming Gao,
Chunlin Wu,
Xue-cheng Tai
Abstract:
$\ell_1$ based sparse regularization plays a central role in compressive sensing and image processing. In this paper, we propose $\ell_1$DecNet, as an unfolded network derived from a variational decomposition model incorporating $\ell_1$ related sparse regularization and solved by scaled alternating direction method of multipliers (ADMM). $\ell_1…
▽ More
$\ell_1$ based sparse regularization plays a central role in compressive sensing and image processing. In this paper, we propose $\ell_1$DecNet, as an unfolded network derived from a variational decomposition model incorporating $\ell_1$ related sparse regularization and solved by scaled alternating direction method of multipliers (ADMM). $\ell_1$DecNet effectively decomposes an input image into a sparse feature and a learned dense feature, and thus helps the subsequent sparse feature related operations. Based on this, we develop $\ell_1$DecNet+, a learnable architecture framework consisting of our $\ell_1$DecNet and a segmentation module which operates over extracted sparse features instead of original images. This architecture combines well the benefits of mathematical modeling and data-driven approaches. To our best knowledge, this is the first study to incorporate mathematical image prior into feature extraction in segmentation network structures. Moreover, our $\ell_1$DecNet+ framework can be easily extended to 3D case. We evaluate the effectiveness of $\ell_1$DecNet+ on two commonly encountered sparse segmentation tasks: retinal vessel segmentation in medical image processing and pavement crack detection in industrial abnormality identification. Experimental results on different datasets demonstrate that, our $\ell_1$DecNet+ architecture with various lightweight segmentation modules can achieve equal or better performance than their enlarged versions respectively. This leads to especially practical advantages on resource-limited devices.
△ Less
Submitted 15 June, 2024; v1 submitted 5 March, 2022;
originally announced March 2022.
-
Branching Brownian motion in a periodic environment and uniqueness of pulsating travelling waves
Authors:
Yan-Xia Ren,
Renming Song,
Fan Yang
Abstract:
Using one-dimensional branching Brownian motion in a periodic environment, we give probabilistic proofs of the asymptotics and uniqueness of pulsating travelling waves of the F-KPP equation in a periodic environment. This paper is a sequel to [Ren et al. Branching Brownian motion in a periodic environment and existence of pulsating travelling waves], in which we proved the existence of the pulsati…
▽ More
Using one-dimensional branching Brownian motion in a periodic environment, we give probabilistic proofs of the asymptotics and uniqueness of pulsating travelling waves of the F-KPP equation in a periodic environment. This paper is a sequel to [Ren et al. Branching Brownian motion in a periodic environment and existence of pulsating travelling waves], in which we proved the existence of the pulsating travelling waves in the supercritical and critical cases using the limits of the additive and derivative martingales of branching Brownian motion in a periodic environment.
△ Less
Submitted 22 February, 2022;
originally announced February 2022.
-
Branching Brownian motion in a periodic environment and existence of pulsating travelling waves
Authors:
Yan-Xia Ren,
Renming Song,
Fan Yang
Abstract:
We study the limits of the additive and derivative martingales of one-dimensional branching Brownian motion in a periodic environment. Then we prove the existence of pulsating travelling wave solutions of the corresponding F-KPP equation in the supercritical and critical cases by representing the solutions probabilistically in terms of the limits of the additive and derivative martingales. We also…
▽ More
We study the limits of the additive and derivative martingales of one-dimensional branching Brownian motion in a periodic environment. Then we prove the existence of pulsating travelling wave solutions of the corresponding F-KPP equation in the supercritical and critical cases by representing the solutions probabilistically in terms of the limits of the additive and derivative martingales. We also prove that there is no pulsating travelling wave solution in the subcritical case. Our main tools are the spine decomposition and martingale change of measures.
△ Less
Submitted 22 February, 2022;
originally announced February 2022.
-
Pseudo Numerical Methods for Diffusion Models on Manifolds
Authors:
Lu** Liu,
Yi Ren,
Zhijie Lin,
Zhou Zhao
Abstract:
Denoising Diffusion Probabilistic Models (DDPMs) can generate high-quality samples such as image and audio samples. However, DDPMs require hundreds to thousands of iterations to produce final samples. Several prior works have successfully accelerated DDPMs through adjusting the variance schedule (e.g., Improved Denoising Diffusion Probabilistic Models) or the denoising equation (e.g., Denoising Di…
▽ More
Denoising Diffusion Probabilistic Models (DDPMs) can generate high-quality samples such as image and audio samples. However, DDPMs require hundreds to thousands of iterations to produce final samples. Several prior works have successfully accelerated DDPMs through adjusting the variance schedule (e.g., Improved Denoising Diffusion Probabilistic Models) or the denoising equation (e.g., Denoising Diffusion Implicit Models (DDIMs)). However, these acceleration methods cannot maintain the quality of samples and even introduce new noise at a high speedup rate, which limit their practicability. To accelerate the inference process while kee** the sample quality, we provide a fresh perspective that DDPMs should be treated as solving differential equations on manifolds. Under such a perspective, we propose pseudo numerical methods for diffusion models (PNDMs). Specifically, we figure out how to solve differential equations on manifolds and show that DDIMs are simple cases of pseudo numerical methods. We change several classical numerical methods to corresponding pseudo numerical methods and find that the pseudo linear multi-step method is the best in most situations. According to our experiments, by directly using pre-trained models on Cifar10, CelebA and LSUN, PNDMs can generate higher quality synthetic images with only 50 steps compared with 1000-step DDIMs (20x speedup), significantly outperform DDIMs with 250 steps (by around 0.4 in FID) and have good generalization on different variance schedules. Our implementation is available at https://github.com/lu**-liu/PNDM.
△ Less
Submitted 31 October, 2022; v1 submitted 20 February, 2022;
originally announced February 2022.
-
Subcritical superprocesses conditioned on non-extinction
Authors:
Rongli Liu,
Yan-Xia Ren,
Renming Song,
Zhenyao Sun
Abstract:
We consider a class of subcritical superprocesses $(X_t)_{t\geq 0}$ with general spatial motions and general branching mechanisms. We study the asymptotic behaviors of $\mathbf Q_{t,r}$, the distribution of $X_t$ conditioned on $X_{t+r}$ not being a null measure. We first give the existence of $\lim_{t\to \infty}\mathbf Q_{t,r}$ and $\lim_{r\to \infty}\mathbf Q_{t,r}$, and then show that an…
▽ More
We consider a class of subcritical superprocesses $(X_t)_{t\geq 0}$ with general spatial motions and general branching mechanisms. We study the asymptotic behaviors of $\mathbf Q_{t,r}$, the distribution of $X_t$ conditioned on $X_{t+r}$ not being a null measure. We first give the existence of $\lim_{t\to \infty}\mathbf Q_{t,r}$ and $\lim_{r\to \infty}\mathbf Q_{t,r}$, and then show that an $L\log L$-type condition is equivalent to the existence of the double limits: $\lim_{r\to \infty} \lim_{t\to\infty}\mathbf Q_{t, r}$ and $\lim_{t\to \infty} \lim_{r\to\infty}\mathbf Q_{t, r}$. Finally, when the $L\log L$-type condition holds, we show that those double limits, and $\lim_{r,t\to \infty}\mathbf Q_{t,r}$, are the same.
△ Less
Submitted 24 October, 2022; v1 submitted 30 December, 2021;
originally announced December 2021.
-
Liouville Theorems for holomorphic maps on pseudo-Hermitian manifolds
Authors:
Haojie Chen,
Yibin Ren
Abstract:
We prove some Liouville type results for generalized holomorphic maps in three classes: maps from pseudo-Hermitian manifolds to almost Hermitian manifolds, maps from almost Hermitian manifolds to pseudo-Hermitian manifolds and maps from pseudo-Hermitian manifolds to pseudo-Hermitian manifolds, assuming that the domains are compact. For instance, we show that any $(J,J^N)$ holomorphic map from a co…
▽ More
We prove some Liouville type results for generalized holomorphic maps in three classes: maps from pseudo-Hermitian manifolds to almost Hermitian manifolds, maps from almost Hermitian manifolds to pseudo-Hermitian manifolds and maps from pseudo-Hermitian manifolds to pseudo-Hermitian manifolds, assuming that the domains are compact. For instance, we show that any $(J,J^N)$ holomorphic map from a compact pseudo-Hermitian manifold $M$ with nonnegative (resp. positive) pseudo-Hermitian sectional curvature to an almost Hermitian manifold $N$ with negative (resp. nonpositive) holomorphic sectional curvature is constant. We also construct explicit almost CR structures on a complex vector bundle over an almost CR manifold.
△ Less
Submitted 7 October, 2021;
originally announced October 2021.
-
Stochastic viscosity solutions of reflected stochastic partial differential equations with non-Lipschitz coefficients
Authors:
Yong Ren,
Jean Marc Owo,
Auguste Aman
Abstract:
This paper, is an attempt to extend the notion of stochastic viscosity solution to reflected semi-linear stochastic partial differential equations (RSPDEs, in short) with non-Lipschitz condition on the coefficients. Our method is fully probabilistic and use the recently developed theory on reflected backward doubly stochastic differential equations (RBDSDEs, in short). Among other, we prove the ex…
▽ More
This paper, is an attempt to extend the notion of stochastic viscosity solution to reflected semi-linear stochastic partial differential equations (RSPDEs, in short) with non-Lipschitz condition on the coefficients. Our method is fully probabilistic and use the recently developed theory on reflected backward doubly stochastic differential equations (RBDSDEs, in short). Among other, we prove the existence of the stochastic viscosity solution, and further extend the nonlinear Feynman-Kac formula to reflected SPDEs, like one appear in \cite{2}. Indeed, in their recent work, Aman and Mrhardy \cite{2} established a stochastic viscosity solution for semi-linear reflected SPDEs with nonlinear Neumann boundary condition by using its connection with RBDSDEs. However, even Aman and Mrhardy consider a general class of reflected SPDEs, all their coefficients are at least Lipschitz. Therefore, our current work can be thought of as a new generalization of a now well-know Feymann-Kac formula to SPDEs with large class of coefficients, which does not seem to exist in the literature. In other words, our work extends (in non boundary case) Aman and Mrhardy's paper.
△ Less
Submitted 2 October, 2021;
originally announced October 2021.
-
Backward stochastic differential equations with time-delayed generators and integrable parameters
Authors:
Auguste Aman,
Yong Ren
Abstract:
In this note, we derive an existence and uniqueness results for delayed backward stochastic differential equation with only integrable data.
In this note, we derive an existence and uniqueness results for delayed backward stochastic differential equation with only integrable data.
△ Less
Submitted 2 October, 2021;
originally announced October 2021.
-
$L^{p}$-solutions of backward stochastic differential equations with time-delayed generators
Authors:
Yong Ren,
Jean Marc Owo,
Auguste Aman
Abstract:
This article is devoted to study the class of backward stochastic differential equation with delayed generator. We suppose the terminal value and the generator to be $L^{p}$-integrable with $p>1$. We derive a new type of estimation related to this BSDE. Next, we establish the existence and uniqueness result in two ways. First, an approximation technics used by Briand et al. (Stochastic Process. Ap…
▽ More
This article is devoted to study the class of backward stochastic differential equation with delayed generator. We suppose the terminal value and the generator to be $L^{p}$-integrable with $p>1$. We derive a new type of estimation related to this BSDE. Next, we establish the existence and uniqueness result in two ways. First, an approximation technics used by Briand et al. (Stochastic Process. Appl. 108 (2003) 109-129) and hence the well-know Picard iterative procedure. Using Picard iterative procedure, we revisit the result of Dos Reis et al. (Stochastic Process. Appl. 121 (9) (2011) 2114-2150), simplifying the proof and give an explicit existence and uniqueness condition related to the Lipschitz constant $K$ and the terminal time $T$.
△ Less
Submitted 2 October, 2021;
originally announced October 2021.
-
Explicit solution for backward stochastic Volterra integral equations with linear time delayed generators
Authors:
Yong Ren,
Harouna Coulibaly,
Auguste Aman
Abstract:
This note aims to give an explicit solution for backward stochastic Volterra integral equations with linear time delayed generators. The process $Y$ is expressed by an integral whose kernel is explicitly given. The processes $Z$ is expressed by Hida-Malliavin derivatives involving $Y$.
This note aims to give an explicit solution for backward stochastic Volterra integral equations with linear time delayed generators. The process $Y$ is expressed by an integral whose kernel is explicitly given. The processes $Z$ is expressed by Hida-Malliavin derivatives involving $Y$.
△ Less
Submitted 2 October, 2021;
originally announced October 2021.
-
Probabilistic representation of parabolic stochastic variational inequality with Dirichlet-Neumann boundary and variational generalized backward doubly stochastic differential equations
Authors:
Yong Ren,
Auguste Aman,
Qing Zhou
Abstract:
We derive the existence and uniqueness of the generalized backward doubly stochastic differential equation with sub-differential of a lower semi-continuous convex function under a non Lipschitz condition. This study allows us give a probabilistic representation (in stochastic viscosity sense) to the parabolic variational stochastic partial differential equations with Dirichlet-Neumann conditions.
We derive the existence and uniqueness of the generalized backward doubly stochastic differential equation with sub-differential of a lower semi-continuous convex function under a non Lipschitz condition. This study allows us give a probabilistic representation (in stochastic viscosity sense) to the parabolic variational stochastic partial differential equations with Dirichlet-Neumann conditions.
△ Less
Submitted 2 October, 2021;
originally announced October 2021.
-
The Seneta-Heyde scaling for supercritical super-Brownian motion
Authors:
Haojie Hou,
Yan-Xia Ren,
Renming Song
Abstract:
We consider the additive martingale $W_t(λ)$ and the derivative martingale $\partial W_t(λ)$ for one-dimensional supercritical super-Brownian motions with general branching mechanism. In the critical case $λ=λ_0$, we prove that $\sqrt{t}W_t(λ_0)$ converges in probability to a positive limit, which is a constant multiple of the almost sure limit $\partial W_\infty(λ_0)$ of the derivative martingale…
▽ More
We consider the additive martingale $W_t(λ)$ and the derivative martingale $\partial W_t(λ)$ for one-dimensional supercritical super-Brownian motions with general branching mechanism. In the critical case $λ=λ_0$, we prove that $\sqrt{t}W_t(λ_0)$ converges in probability to a positive limit, which is a constant multiple of the almost sure limit $\partial W_\infty(λ_0)$ of the derivative martingale $\partial W_t(λ_0)$. We also prove that, on the survival event, $\limsup_{t\to\infty}\sqrt{t}W_t(λ_0)=\infty$ almost surely.
△ Less
Submitted 9 September, 2021;
originally announced September 2021.
-
Convergence rate for a class of supercritical superprocesses
Authors:
Rongli Liu,
Yan-Xia Ren,
Renming Song
Abstract:
Suppose $X=\{X_t, t\ge 0\}$ is a supercritical superprocess. Let $φ$ be the non-negative eigenfunction of the mean semigroup of $X$ corresponding to the principal eigenvalue $λ>0$. Then $M_t(φ)=e^{-λt}\langleφ, X_t\rangle, t\geq 0,$ is a non-negative martingale with almost sure limit $M_\infty(φ)$. In this paper we study the rate at which $M_t(φ)-M_\infty(φ)$ converges to $0$ as $t\to \infty$ when…
▽ More
Suppose $X=\{X_t, t\ge 0\}$ is a supercritical superprocess. Let $φ$ be the non-negative eigenfunction of the mean semigroup of $X$ corresponding to the principal eigenvalue $λ>0$. Then $M_t(φ)=e^{-λt}\langleφ, X_t\rangle, t\geq 0,$ is a non-negative martingale with almost sure limit $M_\infty(φ)$. In this paper we study the rate at which $M_t(φ)-M_\infty(φ)$ converges to $0$ as $t\to \infty$ when the process may not have finite variance. Under some conditions on the mean semigroup, we provide sufficient and necessary conditions for the rate in the almost sure sense. Some results on the convergence rate in $L^p$ with $p\in(1, 2)$ are also obtained.
△ Less
Submitted 14 July, 2021;
originally announced July 2021.
-
Learning particle swarming models from data with Gaussian processes
Authors:
**chao Feng,
Charles Kulick,
Yunxiang Ren,
Sui Tang
Abstract:
Interacting particle or agent systems that display a rich variety of swarming behaviours are ubiquitous in science and engineering. A fundamental and challenging goal is to understand the link between individual interaction rules and swarming. In this paper, we study the data-driven discovery of a second-order particle swarming model that describes the evolution of $N$ particles in $\mathbb{R}^d$…
▽ More
Interacting particle or agent systems that display a rich variety of swarming behaviours are ubiquitous in science and engineering. A fundamental and challenging goal is to understand the link between individual interaction rules and swarming. In this paper, we study the data-driven discovery of a second-order particle swarming model that describes the evolution of $N$ particles in $\mathbb{R}^d$ under radial interactions. We propose a learning approach that models the latent radial interaction function as Gaussian processes, which can simultaneously fulfill two inference goals: one is the nonparametric inference of {the} interaction function with pointwise uncertainty quantification, and the other one is the inference of unknown scalar parameters in the non-collective friction forces of the system. We formulate the learning problem as a statistical inverse problem and provide a detailed analysis of recoverability conditions, establishing that a coercivity condition is sufficient for recoverability. Given data collected from $M$ i.i.d trajectories with independent Gaussian observational noise, we provide a finite-sample analysis, showing that our posterior mean estimator converges in a Reproducing kernel Hilbert space norm, at an optimal rate in $M$ equal to the one in the classical 1-dimensional Kernel Ridge regression. As a byproduct, we show we can obtain a parametric learning rate in $M$ for the posterior marginal variance using $L^{\infty}$ norm, and the rate could also involve $N$ and $L$ (the number of observation time instances for each trajectory), depending on the condition number of the inverse problem. Numerical results on systems that exhibit different swarming behaviors demonstrate efficient learning of our approach from scarce noisy trajectory data.
△ Less
Submitted 7 March, 2023; v1 submitted 4 June, 2021;
originally announced June 2021.
