arXiv subjects

Sean Meyn

Publications and source records attributed to Sean Meyn.

At least 55 records · Page 3Linked to original sources

arXiv2018

Action-Constrained Markov Decision Processes With Kullback-Leibler Cost

This paper concerns computation of optimal policies in which the one-step reward function contains a cost term that models Kullback-Leibler divergence with respect to nominal dynamics. This technique was introduced by Todorov in 2007, where it was shown under general conditions that the solution to the average-reward optimality equations reduce to a simple eigenvector problem. Since then many authors have sought to apply this technique to control problems and models of bounded rationality in economics. A crucial assumption is that the input process is essentially unconstrained. For example, if the nominal dynamics include randomness from nature (e.g., the impact of wind on a moving vehicle), then the optimal control solution does not respect the exogenous nature of this disturbance. This paper introduces a technique to solve a more general class of action-constrained MDPs. The main idea is to solve an entire parameterized family of MDPs, in which the parameter is a scalar weighting the one-step reward function. The approach is new and practical even in the original unconstrained formulation.

math.OC↗

arXiv2016

Ordinary Differential Equation Methods For Markov Decision Processes and Application to Kullback-Leibler Control Cost

A new approach to computation of optimal policies for MDP (Markov decision process) models is introduced. The main idea is to solve not one, but an entire family of MDPs, parameterized by a weighting factor $ζ$ that appears in the one-step reward function. For an MDP with $d$ states, the family of value functions $\{ h^*_ζ: ζ\in\Re\}$ is the solution to an ODE, $$ \frac{d}{dζ} h^*_ζ= {\cal V}(h^*_ζ) $$ where the vector field ${\cal V}\colon\Re^d\to\Re^d$ has a simple form, based on a matrix inverse. This general methodology is applied to a family of average-cost optimal control models in which the one-step reward function is defined by Kullback-Leibler divergence. The motivation for this reward function in prior work is computation: The solution to the MDP can be expressed in terms of the Perron-Frobenius eigenvector for an associated positive matrix. The drawback with this approach is that no hard constraints on the control are permitted. It is shown here that it is possible to extend this framework to model randomness from nature that cannot be modified by the controller. Perron-Frobenius theory is no longer applicable -- the resulting dynamic programming equations appear as complex as a completely unstructured MDP model. Despite this apparent complexity, it is shown that this class of MDPs admits a solution via this new ODE technique. This approach is new and practical even for the simpler problem in which randomness from nature is absent.

math.OC↗

arXiv2016

Estimation and Control of Quality of Service in Demand Dispatch

It is now well known that flexibility of energy consumption can be harnessed for the purposes of grid-level ancillary services. In particular, through distributed control of a collection of loads, a balancing authority regulation signal can be tracked accurately, while ensuring that the quality of service (QoS) for each load is acceptable {\it on average}. In this paper it is argued that a histogram of QoS is approximately Gaussian, and consequently each load will eventually receive poor service. Statistical techniques are developed to estimate the mean and variance of QoS as a function of the power spectral density of the regulation signal. It is also shown that additional local control can eliminate risk: The histogram of QoS is {\it truncated} through this local control, so that strict bounds on service quality are guaranteed. While there is a tradeoff between the grid-level tracking performance (capacity and accuracy) and the bounds imposed on QoS, it is found that the loss of capacity is minor in typical cases.

eess.SY↗

arXiv2016

Approximate optimality with bounded regret in dynamic matching models

We consider a discrete-time bipartite matching model with random arrivals of units of supply and demand that can wait in queues located at the nodes in the network. A control policy determines which are matched at each time. The focus is on the infinite-horizon average-cost optimal control problem. A relaxation of the stochastic control problem is proposed, which is found to be a special case of an inventory model, as treated in the classical theory of Clark and Scarf. The optimal policy for the relaxation admits a closed-form expression. Based on the policy for this relaxation, a new matching policy is proposed. For a parameterized family of models in which the network load approaches capacity, this policy is shown to be approximately optimal, with bounded regret, even though the average cost grows without bound.

cs.DM↗

arXiv2016

Ergodic Theory for Controlled Markov Chains with Stationary Inputs

Consider a stochastic process $\{X(t)\}$ on a finite state space $ {\sf X}=\{1,\dots, d\}$. It is conditionally Markov, given a real-valued `input process' $\{ζ(t)\}$. This is assumed to be small, which is modeled through the scaling, \[ ζ_t = \varepsilon ζ^1_t, \qquad 0\le \varepsilon \le 1\,, \] where $\{ζ^1(t)\}$ is a bounded stationary process. The following conclusions are obtained, subject to smoothness assumptions on the controlled transition matrix and a mixing condition on $\{ζ(t)\}$: (i) A stationary version of the process is constructed, that is coupled with a stationary version of the Markov chain $\{X^\bullet$(t)\}obtained with $\{ζ(t)\}\equiv 0$. The triple $(\{X(t)\}, \{X^\bullet(t)\},\{ζ(t)\})$ is a jointly stationary process satisfying \[ {\sf P}\{X(t) \neq X^\bullet(t)\} = O(\varepsilon) \] Moreover, a second-order Taylor-series approximation is obtained: \[ {\sf P}\{X(t) =i \} ={\sf P}\{X^\bullet(t) =i \} + \varepsilon^2 \varrho(i) + o(\varepsilon^2),\quad 1\le i\le d, \] with an explicit formula for the vector $\varrho\in\mathbb{R}^d$. (ii) For any $m\ge 1$ and any function $f\colon \{1,\dots,d\}\times \mathbb{R}\to\mathbb{R}^m$, the stationary stochastic process $Y(t) = f(X(t),ζ(t))$ has a power spectral density $\text{S}_f$ that admits a second order Taylor series expansion: A function $\text{S}^{(2)}_f\colon [-π,π] \to \mathbb{C}^{ m\times m}$ is constructed such that \[ \text{S}_f(θ) = \text{S}^\bullet_f(θ) + \varepsilon^2 \text{S}_f^{(2)}(θ) + o(\varepsilon^2),\quad θ\in [-π,π] . \] An explicit formula for the function $\text{S}_f^{(2)}$ is obtained, based in part on the bounds in (i). The results are illustrated using a version of the timing channel of Anantharam and Verdu.

cs.PF↗

arXiv2016

State Estimation for the Individual and the Population in Mean Field Control with Application to Demand Dispatch

This paper concerns state estimation problems in a mean field control setting. In a finite population model, the goal is to estimate the joint distribution of the population state and the state of a typical individual. The observation equations are a noisy measurement of the population. The general results are applied to demand dispatch for regulation of the power grid, based on randomized local control algorithms. In prior work by the authors it has been shown that local control can be carefully designed so that the aggregate of loads behaves as a controllable resource with accuracy matching or exceeding traditional sources of frequency regulation. The operational cost is nearly zero in many cases. The information exchange between grid and load is minimal, but it is assumed in the overall control architecture that the aggregate power consumption of loads is available to the grid operator. It is shown that the Kalman filter can be constructed to reduce these communication requirements,

eess.SY↗

arXiv2016

Distributed Randomized Control for Demand Dispatch

The paper concerns design of control systems for Demand Dispatch to obtain ancillary services to the power grid by harnessing inherent flexibility in many loads. The role of "local intelligence" at the load has been advocated in prior work, randomized local controllers that manifest this intelligence are convenient for loads with a finite number of states. The present work introduces two new design techniques for these randomized controllers: (i) The Individual Perspective Design (IPD) is based on the solution to a one-dimensional family of Markov Decision Processes, whose objective function is formulated from the point of view of a single load. The family of dynamic programming equation appears complex, but it is shown that it is obtained through the solution of a single ordinary differential equation. (ii) The System Perspective Design (SPD) is motivated by a single objective of the grid operator: Passivity of any linearization of the aggregate input-output model. A solution is obtained that can again be computed through the solution of a single ordinary differential equation. Numerical results complement these theoretical results.

math.OC↗

arXiv2015

Smart Fridge / Dumb Grid? Demand Dispatch for the Power Grid of 2020

In discussions at the 2015 HICSS meeting, it was argued that loads can provide most of the ancillary services required today and in the future. Through load-level and grid-level control design, high-quality ancillary service for the grid is obtained without impacting quality of service delivered to the consumer. This approach to grid regulation is called demand dispatch: loads are providing service continuously and automatically, without consumer interference. In this paper we ask, what intelligence is required at the grid-level? In particular, does the grid-operator require more than one-way communication to the loads? Our main conclusion: risk is not great in lower frequency ranges, e.g., PJM's RegA or BPA's balancing reserves. In particular, ancillary services from refrigerators and pool-pumps can be obtained successfully with only one-way communication. This requires intelligence at the loads, and much less intelligence at the grid level.

eess.SY↗

arXiv2014

Generalized Error Exponents For Small Sample Universal Hypothesis Testing

The small sample universal hypothesis testing problem is investigated in this paper, in which the number of samples $n$ is smaller than the number of possible outcomes $m$. The goal of this work is to find an appropriate criterion to analyze statistical tests in this setting. A suitable model for analysis is the high-dimensional model in which both $n$ and $m$ increase to infinity, and $n=o(m)$. A new performance criterion based on large deviations analysis is proposed and it generalizes the classical error exponent applicable for large sample problems (in which $m=O(n)$). This generalized error exponent criterion provides insights that are not available from asymptotic consistency or central limit theorem analysis. The following results are established for the uniform null distribution: (i) The best achievable probability of error $P_e$ decays as $P_e=\exp\{-(n^2/m) J (1+o(1))\}$ for some $J>0$. (ii) A class of tests based on separable statistics, including the coincidence-based test, attains the optimal generalized error exponents. (iii) Pearson's chi-square test has a zero generalized error exponent and thus its probability of error is asymptotically larger than the optimal test.

math.ST↗

arXiv2014

Passive Dynamics in Mean Field Control

Mean-field models are a popular tool in a variety of fields. They provide an understanding of the impact of interactions among a large number of particles or people or other "self-interested agents", and are an increasingly popular tool in distributed control. This paper considers a particular randomized distributed control architecture introduced in our own recent work. In numerical results it was found that the associated mean-field model had attractive properties for purposes of control. In particular, when viewed as an input-output system, its linearization was found to be minimum phase. In this paper we take a closer look at the control model. The results are summarized as follows: (i) The Markov Decision Process framework of Todorov is extended to continuous time models, in which the "control cost" is based on relative entropy. This is the basis of the construction of a family of controlled Markovian generators. (ii) A decentralized control architecture is proposed in which each agent evolves as a controlled Markov process. A central authority broadcasts a common control signal to each agent. The central authority chooses this signal based on an aggregate scalar output of the Markovian agents. (iii) Provided the control-free system is a reversible Markov process, the following identity holds for the linearization, \[ \text{Real} (G(jω)) = \text{PSD}_Y(ω)\ge 0, \quad ω\in\Re, \] where the right hand side denotes the power spectral density for the output of any one of the individual (control-free) Markov processes.

eess.SY↗

arXiv2014

Individual risk in mean-field control models for decentralized control, with application to automated demand response

Flexibility of energy consumption can be harnessed for the purposes of ancillary services in a large power grid. In prior work by the authors a randomized control architecture is introduced for individual loads for this purpose. In examples it is shown that the control architecture can be designed so that control of the loads is easy at the grid level: Tracking of a balancing authority reference signal is possible, while ensuring that the quality of service (QoS) for each load is acceptable on average. The analysis was based on a mean field limit (as the number of loads approaches infinity), combined with an LTI-system approximation of the aggregate nonlinear model. This paper examines in depth the issue of individual risk in these systems. The main contributions of the paper are of two kinds: Risk is modeled and quantified: (i) The average performance is not an adequate measure of success. It is found empirically that a histogram of QoS is approximately Gaussian, and consequently each load will eventually receive poor service. (ii) The variance can be estimated from a refinement of the LTI model that includes a white-noise disturbance; variance is a function of the randomized policy, as well as the power spectral density of the reference signal. Additional local control can eliminate risk: (iii) The histogram of QoS is truncated through this local control, so that strict bounds on service quality are guaranteed. (iv) This has insignificant impact on the grid-level performance, beyond a modest reduction in capacity of ancillary service.

math.OC↗

arXiv2014

Ancillary Service to the Grid Using Intelligent Deferrable Loads

Renewable energy sources such as wind and solar power have a high degree of unpredictability and time-variation, which makes balancing demand and supply challenging. One possible way to address this challenge is to harness the inherent flexibility in demand of many types of loads. Introduced in this paper is a technique for decentralized control for automated demand response that can be used by grid operators as ancillary service for maintaining demand-supply balance. A Markovian Decision Process (MDP) model is introduced for an individual load. A randomized control architecture is proposed, motivated by the need for decentralized decision making, and the need to avoid synchronization that can lead to large and detrimental spikes in demand. An aggregate model for a large number of loads is then developed by examining the mean field limit. A key innovation is an LTI-system approximation of the aggregate nonlinear model, with a scalar signal as the input and a measure of the aggregate demand as the output. This makes the approximation particularly convenient for control design at the grid level. The second half of the paper contains a detailed application of these results to a network of residential pools. Simulations are provided to illustrate the accuracy of the approximations and effectiveness of the proposed control approach.

math.OC↗

arXiv2013

Approximate dynamic programming using fluid and diffusion approximations with applications to power management

Neuro-dynamic programming is a class of powerful techniques for approximating the solution to dynamic programming equations. In their most computationally attractive formulations, these techniques provide the approximate solution only within a prescribed finite-dimensional function class. Thus, the question that always arises is how should the function class be chosen? The goal of this paper is to propose an approach using the solutions to associated fluid and diffusion approximations. In order to illustrate this approach, the paper focuses on an application to dynamic speed scaling for power management in computer processors.

cs.LG↗

arXiv2012

Classification with High-Dimensional Sparse Samples

The task of the binary classification problem is to determine which of two distributions has generated a length-$n$ test sequence. The two distributions are unknown; two training sequences of length $N$, one from each distribution, are observed. The distributions share an alphabet of size $m$, which is significantly larger than $n$ and $N$. How does $N,n,m$ affect the probability of classification error? We characterize the achievable error rate in a high-dimensional setting in which $N,n,m$ all tend to infinity, under the assumption that probability of any symbol is $O(m^{-1})$. The results are: 1. There exists an asymptotically consistent classifier if and only if $m=o(\min\{N^2,Nn\})$. This extends the previous consistency result in [1] to the case $N\neq n$. 2. For the sparse sample case where $\max\{n,N\}=o(m)$, finer results are obtained: The best achievable probability of error decays as $-\log(P_e)=J \min\{N^2, Nn\}(1+o(1))/m$ with $J>0$. 3. A weighted coincidence-based classifier has non-zero generalized error exponent $J$. 4. The $\ell_2$-norm based classifier has J=0.

cs.IT↗

arXiv2010

Universal and Composite Hypothesis Testing via Mismatched Divergence

For the universal hypothesis testing problem, where the goal is to decide between the known null hypothesis distribution and some other unknown distribution, Hoeffding proposed a universal test in the nineteen sixties. Hoeffding's universal test statistic can be written in terms of Kullback-Leibler (K-L) divergence between the empirical distribution of the observations and the null hypothesis distribution. In this paper a modification of Hoeffding's test is considered based on a relaxation of the K-L divergence test statistic, referred to as the mismatched divergence. The resulting mismatched test is shown to be a generalized likelihood-ratio test (GLRT) for the case where the alternate distribution lies in a parametric family of the distributions characterized by a finite dimensional parameter, i.e., it is a solution to the corresponding composite hypothesis testing problem. For certain choices of the alternate distribution, it is shown that both the Hoeffding test and the mismatched test have the same asymptotic performance in terms of error exponents. A consequence of this result is that the GLRT is optimal in differentiating a particular distribution from others in an exponential family. It is also shown that the mismatched test has a significant advantage over the Hoeffding test in terms of finite sample size performance. This advantage is due to the difference in the asymptotic variances of the two test statistics under the null hypothesis. In particular, the variance of the K-L divergence grows linearly with the alphabet size, making the test impractical for applications involving large alphabet distributions. The variance of the mismatched divergence on the other hand grows linearly with the dimension of the parameter space, and can hence be controlled through a prudent choice of the function class defining the mismatched divergence.

cs.IT↗

arXiv2010

Feature Extraction for Universal Hypothesis Testing via Rank-constrained Optimization

This paper concerns the construction of tests for universal hypothesis testing problems, in which the alternate hypothesis is poorly modeled and the observation space is large. The mismatched universal test is a feature-based technique for this purpose. In prior work it is shown that its finite-observation performance can be much better than the (optimal) Hoeffding test, and good performance depends crucially on the choice of features. The contributions of this paper include: 1) We obtain bounds on the number of εdistinguishable distributions in an exponential family. 2) This motivates a new framework for feature extraction, cast as a rank-constrained optimization problem. 3) We obtain a gradient-based algorithm to solve the rank-constrained optimization problem and prove its local convergence.

cs.IT↗

arXiv2010

Minimax Robust Quickest Change Detection

The popular criteria of optimality for quickest change detection procedures are the Lorden criterion, the Shiryaev-Roberts-Pollak criterion, and the Bayesian criterion. In this paper a robust version of these quickest change detection problems is considered when the pre-change and post-change distributions are not known exactly but belong to known uncertainty classes of distributions. For uncertainty classes that satisfy a specific condition, it is shown that one can identify least favorable distributions (LFDs) from the uncertainty classes, such that the detection rule designed for the LFDs is optimal for the robust problem in a minimax sense. The condition is similar to that required for the identification of LFDs for the robust hypothesis testing problem originally studied by Huber. An upper bound on the delay incurred by the robust test is also obtained in the asymptotic setting under the Lorden criterion of optimality. This bound quantifies the delay penalty incurred to guarantee robustness. When the LFDs can be identified, the proposed test is easier to implement than the CUSUM test based on the Generalized Likelihood Ratio (GLR) statistic which is a popular approach for such robust change detection problems. The proposed test is also shown to give better performance than the GLR test in simulations for some parameter values.

cs.IT↗

arXiv2008

The ODE method for stability of skip-free Markov chains with applications to MCMC

Fluid limit techniques have become a central tool to analyze queueing networks over the last decade, with applications to performance analysis, simulation and optimization. In this paper, some of these techniques are extended to a general class of skip-free Markov chains. As in the case of queueing models, a fluid approximation is obtained by scaling time, space and the initial condition by a large constant. The resulting fluid limit is the solution of an ordinary differential equation (ODE) in ``most'' of the state space. Stability and finer ergodic properties for the stochastic model then follow from stability of the set of fluid limits. Moreover, similarly to the queueing context where fluid models are routinely used to design control policies, the structure of the limiting ODE in this general setting provides an understanding of the dynamics of the Markov chain. These results are illustrated through application to Markov chain Monte Carlo methods.

math.PR↗