Reinforcement Learning by Policy Search

One objective of artificial intelligence is to model the behavior of an intelligent agent interacting with its environment. The environment's transformations can be modeled as a Markov chain, whose state is partially observable to the agent and affected by its actions; such processes are known as partially observable Markov decision processes (POMDPs). While the environment's dynamics are assumed to obey certain rules, the agent does not know them and must learn. In this dissertation we focus on the agent's adaptation as captured by the reinforcement learning framework. This means learning a policy---a mapping of observations into actions---based on feedback from the environment. The learning can be viewed as browsing a set of policies while evaluating them by trial through interaction with the environment. The set of policies is constrained by the architecture of the agent's controller. POMDPs require a controller to have a memory. We investigate controllers with memory, including controllers with external memory, finite state controllers and distributed controllers for multi-agent systems. For these various controllers we work out the details of the algorithms which learn by ascending the gradient of expected cumulative reinforcement. Building on statistical learning theory and experiment design theory, a policy evaluation algorithm is developed for the case of experience re-use. We address the question of sufficient experience for uniform convergence of policy evaluation and obtain sample complexity bounds for various estimators. Finally, we demonstrate the performance of the proposed algorithms on several domains, the most complex of which is simulated adaptive packet routing in a telecommunication network.

Logistics Network Design with Differentiated Delivery Lead-Time: Benefits and Insights

Most logistics network design models assume exogenous customer demand that is independent of the service time or level. This paper examines the benefits of segmenting demand according to lead-time sensitivity of customers. To capture lead-time sensitivity in the network design model, we use a facility grouping method to ensure that the different demand classes are satisfied on time. In addition, we perform a series of computational experiments to develop a set of managerial insights for the network design decision making process.

Static Pricing for a Network Service Provider

This article studies the static pricing problem of a network service provider who has a fixed capacity and faces different types of customers (classes). Each type of customers can have its own capacity constraint but it is assumed that all classes have the same resource requirement. The provider must decide a static price for each class. The customer types are characterized by their arrival process, with a price-dependant arrival rate, and the random time they remain in the system. Many real-life situations could fit in this framework, for example an Internet provider or a call center, but originally this problem was thought for a company that sells phone-cards and needs to set the price-per-minute for each destination. Our goal is to characterize the optimal static prices in order to maximize the provider's revenue. We note that the model here presented, with some slight modifications and additional assumptions can be used in those cases when the objective is to maximize social welfare.

The Complexity of Safety Stock Placement in General-Network Supply Chains

We consider the optimization problem of safety stock placement in a supply chain, as formulated in [1]. We prove that this problem is NP-Hard for supply chains modeled as general acyclic networks. Thus, we do not expect to find a polynomial-time algorithm for safety stock placement for a general-network supply chain.