Patrick R. Jordan

Empirical game-theoretic analysis of the TAC Supply Chain game

The TAC Supply Chain Management (TAC/SCM) game presents a challenging dynamic environment for autonomous decision-making in a salient application domain. Strategic interactions complicate the analysis of games such as TAC/SCM. since the effectiveness of a given strategy depends on the strategies played by other agents on the supply chain. The TAC tournament generates results from one particular path of combinations, and success in the tournament is rightly regarded as evidence for agent quality. Such results along with post-competition controlled experiments provide useful evaluations of novel techniques employed in the game. We argue that a broader game-theoretic analysis framework can provide a firmer foundation for choice of experimental contexts. Exploiting a repository of agents from the 2005 and 2006 TAC/SCM tournaments, we demonstrate an empirical game-theoretic methodology based on extensive simulation and careful measurement. Our analysis of agents from TAC-05 reveals interesting interactions not seen in the tournament. Extending the analysis to TAC-06 enables us to measure progress from year-to-year, and generates a candidate empirical equilibrium among the best known strategies. We use this equilibrium as a stable background population for comparing relative performance of the 2006 agents, yielding insights complementing the tournament results.

Controlling a supply chain agent using value-based decomposition

We present and evaluate the design of Deep Maize, our entry in the 2005 Trading Agent Competition Supply Chain Management scenario. The central idea is to decompose the problem by estimating the value of key resources in the game. We first create a high-level production schedule that considers cross-cutting constraints and future decisions, but abstracts aways from the details of sales and purchasing. We then make specific sales and purchasing decisions separately, coordinating these decisions with the high-level schedule using resource values derived from the schedule. All of these decisions are made using approximate optimization techniques and make use of explicit predictions about market conditions. Deep Maize was one of the most successful agents in the 2005 tournament, both in overall performance and on specific measures that emphasize coordination.

Forecasting market prices in a supply chain game

Future market conditions can be a pivotal factor in making business decisions. We present and evaluate methods used by our agent, Deep Maize, to forecast market prices in the Trading Agent Competition Supply Chain Management Game. As a guiding principle we seek to exploit as many sources of available information as possible to inform predictions. Since information comes in several different forms, we integrate well-known methods in a novel way to make predictions. The core of our predictor is a nearest-neighbors machine learning algorithm that identifies historical instances with similar economic indicators. We augment this with an online learning procedure that transforms the predictions by optimizing a scoring rule. This allows us to select more relevant historical contexts using additional information available during individual games. We also explore the advantages of two different representations for predicting price distributions. One uses absolute prices, and the other uses statistics of price time series to exploit local stability. We evaluate these methods using both data from the 2005 tournament final round and additional simulations. We compare several variations of our predictor to one another and a baseline predictor similar to those used by many other tournament agents. We show substantial improvements over the baseline predictor, and demonstrate that each element of our predictor contributes to improved performance.

Designing an Ad Auctions Game for the Trading Agent Competition

We introduce the TAC Ad Auctions game (TAC/AA), a new game for the Trading Agent Competition. The Ad Auctions game investigates complex strategic issues found in real sponsored search auctions that are not captured in current analytical models. We provide an overview of TAC/AA, introducing its key features and design rationale. TAC/AA debuted in summer 2009, with the final tournament commencing in conjunction with the TADA-09 workshop.

LexNet: A Graphical Environment for Graph-Based NLP

This interactive presentation describes LexNet, a graphical environment for graph-based NLP developed at the University of Michigan. LexNet includes LexRank (for text summarization), biased LexRank (for passage retrieval), and TUMBL (for binary classification). All tools in the collection are based on random walks on lexical graphs, that is graphs where different NLP objects (e.g., sentences or phrases) are represented as nodes linked by edges proportional to the lexical similarity between the two nodes. We will demonstrate these tools on a variety of NLP tasks including summarization, question answering, and prepositional phrase attachment.