arXiv subjects

Hao Shen

Publications and source records attributed to Hao Shen.

At least 127 records · Page 7Linked to original sources

arXiv2018

The dynamical sine-Gordon model in the full subcritical regime

We prove that the dynamical sine-Gordon equation on the two dimensional torus introduced in [HS16] is locally well-posed for the entire subcritical regime. At first glance this equation is far out of the scope of the local existence theory available in the framework of regularity structures [Hai14, BHZ16, CH16, BCCH17] since it involves a non-polynomial nonlinearity and the solution is expected to be a distribution (without any additional small parameter as in [FG17, HX18]). In [HS16] this was overcome by a change of variable, but the new equation that arises has a multiplicative dependence on highly non-Gaussian noises which makes stochastic estimates highly non-trivial - as a result [HS16] was only able to treat part of the subcritical regime. Moreover, the cumulants of these noises fall out of the scope of the later work [CH16]. In this work we systematically leverage "charge" cancellations specific to this model and obtain stochastic estimates that allow us to cover the entire subcritical regime.

math.PR↗

arXiv2018

Stochastic Telegraph Equation Limit for the Stochastic Six Vertex Model

In this article we study the stochastic six vertex model under the scaling proposed by Borodin and Gorin (2018), where the weights of corner-shape vertices are tuned to zero, and prove Conjecture 6.1 therein: that the height fluctuation converges in finite dimensional distributions to the solution of stochastic telegraph equation.

math.PR↗

arXiv2018

Open ASEP in the Weakly Asymmetric Regime

We consider ASEP on a bounded interval and on a half line with sources and sinks. On the full line, Bertini and Giacomin proved convergence under weakly asymmetric scaling of the height function to the solution of the KPZ equation. We prove here that under similar weakly asymmetric scaling of the sources and sinks as well, the bounded interval ASEP height function converges to the KPZ equation on the unit interval with Neumann boundary conditions on both sides (different parameter for each side); and likewise for the half line ASEP to KPZ on a half line. This result can be interpreted as showing that the KPZ equation arises at the triple critical point (maximal current / high density / low density) of the open ASEP.

math.PR↗

arXiv2018

Glauber dynamics of 2D Kac-Blume-Capel model and their stochastic PDE limits

We study the Glauber dynamics of a two dimensional Blume-Capel model (or dilute Ising model) with Kac potential parametrized by $(β,θ)$ - the "inverse temperature" and the "chemical potential". We prove that the locally averaged spin field rescales to the solution of the dynamical $Φ^4$ equation near a curve in the $(β,θ)$ plane and to the solution of the dynamical $Φ^6$ equation near one point on this curve. Our proof relies on a discrete implementation of Da Prato-Debussche method as in a result by Mourrat-Weber but an additional coupling argument is needed to show convergence of the linearized dynamics.

math.PR↗

arXiv2017

Towards a Mathematical Understanding of the Difficulty in Learning with Feedforward Neural Networks

Training deep neural networks for solving machine learning problems is one great challenge in the field, mainly due to its associated optimisation problem being highly non-convex. Recent developments have suggested that many training algorithms do not suffer from undesired local minima under certain scenario, and consequently led to great efforts in pursuing mathematical explanations for such observations. This work provides an alternative mathematical understanding of the challenge from a smooth optimisation perspective. By assuming exact learning of finite samples, sufficient conditions are identified via a critical point analysis to ensure any local minimum to be globally minimal as well. Furthermore, a state of the art algorithm, known as the Generalised Gauss-Newton (GGN) algorithm, is rigorously revisited as an approximate Newton's algorithm, which shares the property of being locally quadratically convergent to a global minimum under the condition of exact learning.

cs.LG↗

arXiv2017

Moment bounds for SPDEs with non-Gaussian fields and application to the Wong-Zakai problem

Upon its inception the theory of regularity structures allowed for the treatment for many semilinear perturbations of the stochastic heat equation driven by space-time white noise. When the driving noise is non-Gaussian the machinery of theory can still be used but must be combined with an infinite number of stochastic estimates in order to compensate for the loss of hypercontractivity. In this paper we obtain a more streamlined and automatic set of criteria implying these estimates which facilitates the treatment of some other problems including non-Gaussian noise such as some general phase coexistence models - as an example we prove here a generalization of the Wong-Zakai Theorem found by Hairer and Pardoux.

math.PR↗

arXiv2016

ASEP(q,j) converges to the KPZ equation

We show that a generalized Asymmetric Exclusion Process called ASEP(q,j) introduced by Carinci, Giardina, Redig and Sasamoto converges to the Cole-Hopf solution to the KPZ equation under weak asymmetry scaling.

math.PR↗

arXiv2016

Reinforcement Learning in Conflicting Environments for Autonomous Vehicles

In this work, we investigate the application of Reinforcement Learning to two well known decision dilemmas, namely Newcomb's Problem and Prisoner's Dilemma. These problems are exemplary for dilemmas that autonomous agents are faced with when interacting with humans. Furthermore, we argue that a Newcomb-like formulation is more adequate in the human-machine interaction case and demonstrate empirically that the unmodified Reinforcement Learning algorithms end up with the well known maximum expected utility solution.

cs.AI↗

arXiv2016

A central limit theorem for the KPZ equation

We consider the KPZ equation in one space dimension driven by a stationary centred space-time random field, which is sufficiently integrable and mixing, but not necessarily Gaussian. We show that, in the weakly asymmetric regime, the solution to this equation considered at a suitable large scale and in a suitable reference frame converges to the Hopf-Cole solution to the KPZ equation driven by space-time Gaussian white noise. While the limiting process depends only on the integrated variance of the driving field, the diverging constants appearing in the definition of the reference frame also depend on higher order moments.

math.PR↗

arXiv2016

$\ell_1$ Regularized Gradient Temporal-Difference Learning

In this paper, we study the Temporal Difference (TD) learning with linear value function approximation. It is well known that most TD learning algorithms are unstable with linear function approximation and off-policy learning. Recent development of Gradient TD (GTD) algorithms has addressed this problem successfully. However, the success of GTD algorithms requires a set of well chosen features, which are not always available. When the number of features is huge, the GTD algorithms might face the problem of overfitting and being computationally expensive. To cope with this difficulty, regularization techniques, in particular $\ell_1$ regularization, have attracted significant attentions in developing TD learning algorithms. The present work combines the GTD algorithms with $\ell_1$ regularization. We propose a family of $\ell_1$ regularized GTD algorithms, which employ the well known soft thresholding operator. We investigate convergence properties of the proposed algorithms, and depict their performance with several numerical experiments.

cs.AI↗

arXiv2016

Monitoring and Prediction in Smart Energy Systems via Multi-timescale Nexting

Reliable prediction of system status is a highly demanded functionality of smart energy systems, which can enable users or human operators to react quickly to potential future system changes. By adopting the multi-timescale nexting method, we develop an architecture of human-in-the-loop energy control system, which is capable of casting short-term predictive information about the specific smart energy system. The developed architecture does either require a system model nor additional acquisition of (sensor) data in the existing system configuration. Our first experiments demonstrate the performance of the proposed control architecture in an electrical heating system simulation. In the second experiment, we verify the effectiveness of our developed structure in simulating a heating system in a thermal model of a building, by employing natural EnergyPlus temperature data.

eess.SY↗

arXiv2016

Optimal Water Heater Control in Smart Home Environments

In this work, we develop an optimal water heater control method for a smart home environment. It is important to notice that unlike battery storage systems, energy flow in a water heater control system is not reversible. In order to increase the eigen consumption of photovoltaic energy, i.e. direct consumption of generated energy in the house, we propose to employ a dynamic programming approach to optimize heating schedules using forecasted consumption and weather data. Simulation results demonstrate the capability of our proposed system in reducing the overall energy cost while maintaining residents' comfort.

eess.SY↗

arXiv2016

Weak universality of dynamical $Φ^4_3$: non-Gaussian noise

We consider a class of continuous phase coexistence models in three spatial dimensions. The fluctuations are driven by symmetric stationary random fields with sufficient integrability and mixing conditions, but not necessarily Gaussian. We show that, in the weakly nonlinear regime, if the external potential is a symmetric polynomial and a certain average of it exhibits pitchfork bifurcation, then these models all rescale to $Φ^4_3$ near their critical point.

math-ph↗

arXiv2015

The strict-weak lattice polymer

We introduce the strict-weak polymer model, and show the KPZ universality of the free energy fluctuation of this model for a certain range of parameters. Our proof relies on the observation that the discrete time geometric q-TASEP model, studied earlier by A. Borodin and I. Corwin, scales to this polymer model in the limit q->1. This allows us to exploit the exact results for geometric q-TASEP to derive a Fredholm determinant formula for the strict-weak polymer, and in turn perform rigorous asymptotic analysis to show KPZ scaling and GUE Tracy-Widom limit for the free energy fluctuations. We also derive moments formulae for the polymer partition function directly by Bethe ansatz, and identify the limit of the free energy using a stationary version of the polymer model.

math.PR↗

arXiv2015

Subsampled terahertz data reconstruction based on spatio-temporal dictionary learning

In this paper, the problem of terahertz pulsed imaging and reconstruction is addressed. It is assumed that an incomplete (subsampled) three dimensional THz data set has been acquired and the aim is to recover all missing samples. A sparsity-inducing approach is proposed for this purpose. First, a simple interpolation is applied to incomplete noisy data. Then, we propose a spatio-temporal dictionary learning method to obtain an appropriate sparse representation of data based on a joint sparse recovery algorithm. Then, using the sparse coefficients and the learned dictionary, the 3D data is effectively denoised by minimizing a simple cost function. We consider two types of terahertz data to evaluate the performance of the proposed approach; THz data acquired for a model sample with clear layered structures (e.g., a T-shape plastic sheet buried in a polythene pellet), and pharmaceutical tablet data (with low spatial resolution). The achieved signal-to-noise-ratio for reconstruction of T-shape data, from only 5% observation was 19 dB. Moreover, the accuracies of obtained thickness and depth measurements for pharmaceutical tablet data after reconstruction from 10% observation were 98.8%, and 99.9%, respectively. These results, along with chemical mapping analysis, presented at the end of this paper, confirm the accuracy of the proposed method.

cs.CV↗

arXiv2015

Texture Retrieval via the Scattering Transform

This work studies the problem of content-based image retrieval, specifically, texture retrieval. It focuses on feature extraction and similarity measure for texture images. Our approach employs a recently developed method, the so-called Scattering transform, for the process of feature extraction in texture retrieval. It shares a distinctive property of providing a robust representation, which is stable with respect to spatial deformations. Recent work has demonstrated its capability for texture classification, and hence as a promising candidate for the problem of texture retrieval. Moreover, we adopt a common approach of measuring the similarity of textures by comparing the subband histograms of a filterbank transform. To this end we derive a similarity measure based on the popular Bhattacharyya Kernel. Despite the popularity of describing histograms using parametrized probability density functions, such as the Generalized Gaussian Distribution, it is unfortunately not applicable for describing most of the Scattering transform subbands, due to the complex modulus performed on each one of them. In this work, we propose to use the Weibull distribution to model the Scattering subbands of descendant layers. Our numerical experiments demonstrated the effectiveness of the proposed approach, in comparison with several state of the arts.

cs.IR↗

arXiv2014

A renormalization group method by harmonic extensions and the classical dipole gas

In this paper we develop a new renormalization group method, which is based on conditional expectations and harmonic extensions, to study functional integrals related with small perturbations of Gaussian fields. In this new method one integrates Gaussian fields inside domains at all scales conditioning on the fields outside these domains, and by variation principle solves local elliptic problems. It does not rely on an a priori decomposition of the Gaussian covariance. We apply this method to the model of classical dipole gas on the lattice, and show that the scaling limit of the generating function with smooth test functions is the generating function of the renormalized Gaussian free field.

math-ph↗

arXiv2014

The dynamical sine-Gordon model

We introduce the dynamical sine-Gordon equation in two space dimensions with parameter $β$, which is the natural dynamic associated to the usual quantum sine-Gordon model. It is shown that when $β^2 \in (0,\frac{16π}{3})$ the Wick renormalised equation is well-posed. In the regime $β^2 \in (0,4π)$, the Da Prato-Debussche method applies, while for $β^2 \in [4π,\frac{16π}{3})$, the solution theory is provided via the theory of regularity structures (Hairer 2013). We also show that this model arises naturally from a class of $2+1$-dimensional equilibrium interface fluctuation models with periodic nonlinearities. The main mathematical difficulty arises in the construction of the model for the associated regularity structure where the role of the noise is played by a non-Gaussian random distribution similar to the complex multiplicative Gaussian chaos recently analysed by Lacoin, Rhodes and Vargas (2013).

math.PR↗