arXiv subjects

Tatiana Tatarenko

Publications and source records attributed to Tatiana Tatarenko.

29 records · Page 2Linked to original sources

arXiv2020

Minimizing Regret of Bandit Online Optimization in Unconstrained Action Spaces

We consider online convex optimization with a zero-order oracle feedback. In particular, the decision maker does not know the explicit representation of the time-varying cost functions, or their gradients. At each time step, she observes the value of the corresponding cost function evaluated at her chosen action (zero-order oracle). The objective is to minimize the regret, that is, the difference between the sum of the costs she accumulates and that of a static optimal action had she known the sequence of cost functions a priori. We present a novel algorithm to minimize regret in unconstrained action spaces. Our algorithm hinges on a classical idea of one-point estimation of the gradients of the cost functions based on their observed values. The algorithm is independent of problem parameters. Letting $T$ denote the number of queries of the zero-order oracle and $n$ the problem dimension, the regret rate achieved is $O(n^{2/3}T^{2/3})$. Moreover, we adapt the presented algorithm to the setting with two-point feedback and demonstrate that the adapted procedure achieves the theoretical lower bound on the regret of $(n^{1/2}T^{1/2})$.

math.OC↗

arXiv2020

A Smooth Inexact Penalty Reformulation of Convex Problems with Linear Constraints

In this work, we consider a constrained convex problem with linear inequalities and provide an inexact penalty re-formulation of the problem. The novelty is in the choice of the penalty functions, which are smooth and can induce a non-zero penalty over some points in feasible region of the original constrained problem. The resulting unconstrained penalized problem is parametrized by two penalty parameters which control the slope and the curvature of the penalty function. With a suitable selection of these penalty parameters, we show that the solutions of the resulting penalized unconstrained problem are \emph{feasible} for the original constrained problem, under some assumptions. Also, we establish that, with suitable choices of penalty parameters, the solutions of the penalized unconstrained problem can achieve a suboptimal value which is arbitrarily close to the optimal value of the original constrained problem. For the problems with a large number of linear inequality constraints, a particular advantage of such a smooth penalty-based reformulation is that it renders a penalized problem suitable for the implementation of fast incremental gradient methods, which require only one sample from the inequality constraints at each iteration. We consider applying SAGA proposed in \cite{saga} to solve the resulting penalized unconstrained problem. Moreover, we propose an alternative approach to set up the penalized problem. This approach is based on the time-varying penalty parameters and, thus, does not require knowledge about some problem-specific properties, that might be difficult to estimate. We prove that the single-loop full gradient-based algorithm applied to the corresponding time-varying penalized problem converges to the solution of the original constrained problem in the case of the strongly convex objective function.

math.OC↗

arXiv2020

Convergence Rate of a Penalty Method for Strongly Convex Problems with Linear Constraints

We consider an optimization problem with strongly convex objective and linear inequalities constraints. To be able to deal with a large number of constraints we provide a penalty reformulation of the problem. As penalty functions we use a version of the one-sided Huber losses. The smoothness properties of these functions allow us to choose time-varying penalty parameters in such a way that the incremental procedure with the diminishing step-size converges to the exact solution with the rate $O(1/{\sqrt k})$. To the best of our knowledge, we present the first result on the convergence rate for the penalty-based gradient method, in which the penalty parameters vary with time.

math.OC↗

arXiv2019

Geometric Convergence of Distributed Gradient Play in Games with Unconstrained Action Sets

We provide a distributed algorithm to learn a Nash equilibrium in a class of non-cooperative games with strongly monotone mappings and unconstrained action sets. Each player has access to her own smooth local cost function and can communicate to her neighbors in some undirected graph. We consider a distributed communication-based gradient algorithm. For this procedure, we prove geometric convergence to a Nash equilibrium. In contrast to our previous works [15], [16], where the proposed algorithms required two parameters to be set up and the analysis was based on a so called augmented game mapping, the procedure in this work corresponds to a standard distributed gradient play and, thus, only one constant step size parameter needs to be chosen appropriately to guarantee fast convergence to a game solution. Moreover, we provide a rigorous comparison between the convergence rate of the proposed distributed gradient play and the rate of the GRANE algorithm presented in [15]. It allows us to demonstrate that the distributed gradient play outperforms the GRANE in terms of convergence speed.

math.OC↗

arXiv2019

Penalized Push-Sum Algorithm for Constrained Distributed Optimization with Application to Energy Management in Smart Grid

We study distributed convex constrained optimization on a time-varying multi-agent network. Each agent has access to its own local cost function, its local constraints, and its instant number of out-neighbors. The collective goal is to minimize the sum of the cost functions over the set of all constraints. We utilize the push-sum protocol to be able to solve this distributed optimization problem. We adapt the push-sum optimization algorithm, which has been studied in context of unconstrained optimization so far, to convex constrained optimization by introducing an appropriate choice of penalty functions and penalty parameters. Under some additional technical assumptions on the gradients we prove convergence of the distributed penalty-based push-sum algorithm to the optimal value of the global objective function. We apply the proposed penalty-based push-sum algorithm to the problem of distributed energy management in smart grid and discuss the advantages of this novel procedure in comparison with existing ones.

eess.SP↗

arXiv2019

Learning Nash Equilibria in Monotone Games

We consider multi-agent decision making where each agent's cost function depends on all agents' strategies. We propose a distributed algorithm to learn a Nash equilibrium, whereby each agent uses only obtained values of her cost function at each joint played action, lacking any information of the functional form of her cost or other agents' costs or strategy sets. In contrast to past work where convergent algorithms required strong monotonicity, we prove algorithm convergence under mere monotonicity assumption. This significantly widens algorithm's applicability, such as to games with linear coupling constraints.

cs.MA↗

arXiv2018

Geometric Convergence of Gradient Play Algorithms for Distributed Nash Equilibrium Seeking

We study distributed algorithms for seeking a Nash equilibrium in a class of non-cooperative convex games with strongly monotone mappings. Each player has access to her own smooth local cost function and can communicate to her neighbors in some undirected graph. To deal with fast distributed learning of Nash equilibria under such settings, we introduce a so called augmented game mapping and provide conditions under which this mapping is strongly monotone. We consider a distributed gradient play algorithm for determining a Nash equilibrium (GRANE). The algorithm involves every player performing a gradient step to minimize her own cost function while sharing and retrieving information locally among her neighbors in the network. Using the reformulation of the Nash equilibrium problem based on the strong monotone augmented game mapping, we prove the convergence of this algorithm to a Nash equilibrium with a geometric rate. Further, we introduce the Nesterov type acceleration for the gradient play algorithm. We demonstrate that, similarly to the accelerated algorithms in centralized optimization and variational inequality problems, our accelerated algorithm outperforms GRANE in the convergence rate. Moreover, to relax assumptions required to guarantee the strongly monotone augmented mapping, we analyze the restricted strongly monotone property of this mapping and prove geometric convergence of the distributed gradient play under milder assumptions.

math.OC↗

arXiv2018

Learning Generalized Nash Equilibria in a Class of Convex Games

We consider multi-agent decision making where each agent optimizes its convex cost function subject to individual and coupling constraints. The constraint sets are compact convex subsets of a Euclidean space. To learn Nash equilibria, we propose a novel distributed payoff-based algorithm, such that each agent uses information only about its cost function values and the constraint function values with their associated dual multiplier. We prove convergence of this algorithm to a Nash equilibrium, under the assumption that the game admits a strictly convex potential function. In the absence of coupling constraints, we prove convergence to Nash equilibria under significantly weaker assumptions, not requiring a potential function. Namely, strict monotonicity of the game mapping is sufficient for convergence. We also derive the convergence rate of the algorithm for strongly monotone game maps.

math.OC↗

arXiv2018

Stochastic Learning in Potential Games: Communication and Payoff-Based Approaches

Game theory serves as a powerful tool for distributed optimization in multi-agent systems in different applications. In this paper we consider multi-agent systems that can be modeled by means of potential games whose potential function coincides with a global objective function to be maximized. In this approach, the agents correspond to the strategic decision makers and the optimization problem is equivalent to the problem of learning a potential function maximizer in the designed game. The paper deals with two different information settings in the system. Firstly, we consider systems, where agents have the access to the gradient of their utility functions. However, they do not possess the full information about the joint actions. Thus, to be able to move along the gradient toward a local optimum, they need to exchange the information with their neighbors by means of communication. The second setting refers to a payoff-based approach to learning potential function maximizers. Here, we assume that at each iteration agents can only observe their own played actions and experienced payoffs. In both cases, the paper studies unconstrained non-convex optimization with a differentiable objective function. To develop the corresponding algorithms guaranteeing convergence to a local maximum of the potential function in the game, we utilize the idea of the well-known Robbins-Monro procedure based on the theory of stochastic approximation.

math.OC↗

arXiv2016

Non-Convex Distributed Optimization

We study distributed non-convex optimization on a time-varying multi-agent network. Each node has access to its own smooth local cost function, and the collective goal is to minimize the sum of these functions. We generalize the results obtained previously to the case of non-convex functions. Under some additional technical assumptions on the gradients we prove the convergence of the distributed push-sum algorithm to some critical point of the objective function. By utilizing perturbations on the update process, we show the almost sure convergence of the perturbed dynamics to a local minimum of the global objective function. Our analysis shows that this noised procedure converges at a rate of $O(1/t)$.

math.OC↗

arXiv2016

Payoff-Based Approach to Learning Nash Equilibria in Convex Games

We consider multi-agent decision making, where each agent optimizes its cost function subject to constraints. Agents' actions belong to a compact convex Euclidean space and the agents' cost functions are coupled. We propose a distributed payoff-based algorithm to learn Nash equilibria in the game between agents. Each agent uses only information about its current cost value to compute its next action. We prove convergence of the proposed algorithm to a Nash equilibrium in the game leveraging established results on stochastic processes. The performance of the algorithm is analyzed with a numerical case study.

math.OC↗