Ioannis A. Vetsikas' webpage

 

About Me

I am a Senior Research Fellow in the Department of Electronics and Computer Science of the University of Southampton (UK). I work with prof. Nick Jennings.

I got my Ph.D. from the Department of Computer Science of Cornell University (USA) under the supervision of prof. Bart Selman.

I got my Diploma (5 year degree with thesis) in Electrical and Computer Engineering, with focus on Computer Engineering and Sciences at the National Technical Univ. of Athens, Greece.

 

Research Statement

My primary, long-term research goal is to create the methodology and theory necessary for software agents to effectively participate in real-world online auctions. The proliferation of such auctions on the internet has opened them up to millions of private individuals, who do not necessarily possess the professional expertise to determine how they should bid to maximize their return. Given this, the agents that I design aim to assist and represent their owners in these auctions. To achieve this, however, these agents should account for the features of real-world auctions that expert bidders take into consideration when determining their bidding strategies. Traditional auction theory models do not account for all important features of real-world auctions; this is the reason why I am both extending the fundamental auction theory in this area and making it applicable to realistic settings. In particular, I am looking at auctions with reserve prices and agents with budget constraints, which can have any risk attitude. Additionally, agents can be competitive and try to outperform their competitors, rather than simply being self-interested, and have a desire to buy multiple goods. They can participate in multiple auctions, which take place sequentially and concurrently, and they can have uncertain knowledge about: the number of opponents faced and of goods being sold, the value of the goods they are buying, the closing time of the auction, the opponents’ competitiveness and risk attitudes. While some of these issues have been examined individually for single-unit sealed-bid auctions, they have not been combined together; my research addresses this issue by examining all these features, primarily within the context of multi-unit sealed-bid auctions.

Where it is not possible to provide a purely theoretical analysis, I have expanded into looking at methodologies for designing agents, which combine both theoretical and empirical knowledge. Furthermore, I am also interested in the design of auctions and mechanisms. These can range from using the results of the analysis on the individual bidding strategies to determine the optimal auction from an auctioneer’s point of view, to new auction formats that can be applied within the internet auctions available today, all the way to novel mechanisms that can achieve goals such as efficiency and fairness.

Additionally, I’m looking into the application of auction mechanism is various problems, which impact significantly on the everyday life of people and companies within the current digital economy. Some examples are resource allocation, when there are constraints such as limited communication, which is implemented primarily in the context of disaster management scenarios, scenarios where auctions are used for service provisioning, such as acquiring resources for ubiquitous computing or other everyday services, and use of auctions in energy settings, such as energy markets or smart meters for homes that use agent technologies to reduce the energy demand and costs.

 

I'm also involved in the Trading Agent Competition, and my entry agent WhiteBear is the most consistently high-scoring agent in the history of the TAC Classic game. Our new agent IAMwildCAT is the winner of the new TAC CAT game in 2008.

 

Publications

2010

  • Bidding strategies for realistic multi-unit sealed-bid auctions [pdf]
    I. A. Vetsikas and N. R. Jennings
    Journal of Autonomous Agents and Multi-Agent Systems (JAAMAS), 21(2), 265-291, DOI : 10.1007/s10458-009-9109-6. (Published first online in October 2009)
    This paper extends the work of [AAAI08]. In addition to the work presented there, we examine what happens when valuation uncertainty exists in an m-th price setting and have a number of new experiments and more analysis in general.
  • Sequential Auctions With Partially Substitutable Goods [pdf]
    I. A. Vetsikas and N. R. Jennings
    Book Chapter in Selected Papers from AMEC/TADA 09, Springer LNBIP 59, 242-258.
    In this paper we present the analysis of how to bid when you want to purchase one of a sequence of partially substitutable items.
    Also appeared in: IJCAI Workshop on Trading Agent Design & Analysis (TADA), Los Angeles, USA, 123-130, (2009).
  • Designing Trading Agents for Real-world Auctions [pdf]
    I. A. Vetsikas and N. R. Jennings
    Proc. 6th Hellenic Conference on Artificial Intelligence (SETN 2010), Athens, Greece, LNCS Vol. 6040, 275-285.
    This paper presents my personal agenda for the research I have been doing and where it is going. It unifies much of the work presented in many of my other papers.
  • Flexibly priced options: A new mechanism for sequential auction markets with complementary goods [ext abs] [pdf]
    V. Robu, I. A. Vetsikas, E. Gerding and N. R. Jennings
    Proc. 9th International Joint Conference on Autonomous Agents and Multi-agent Systems (AAMAS), Toronto, Canada, 1485-1486.
    In this paper, we present flexibly priced options, which determine both the option and exercise price from the bids placed by the buyers.
    Also appeared in: Agent Mediated Electronic Commerce (AMEC) Workshop in AAMAS-10, 29-42, (2010). [pdf]
  • Multi-unit Auctions with Asymmetric Bidders [pdf]
    I. A. Vetsikas and N. R. Jennings
    Proc. ECAI Workshop on Intelligent Agent Technologies for e-Business (IAT4EB), Lisbon, Portugal, 1-6.
    This paper extends and corrects [MBC08]. In addition to the old model, where it was assumed that the models of the opponents were not known, in this paper we also examine the case when this information is common knowledge.
  • Addressing the Exposure Problem of Bidding Agents Using Flexibly Priced Options [pdf]
    V. Robu, I. A. Vetsikas, E. Gerding and N. R. Jennings
    Proc. 19th European Conference on Artificial Intelligence (ECAI), Lisbon, Portugal, 581-586.
    This paper extends [AMEC-AAMAS10]. In addition to the old model, we introduce generalized flexibly priced options. These determine both the option and exercise price from the bids placed by the options, and generalize and extend all the option models that are available in the literature. We show that they are the way to maximize both the seller profit and the allocative efficiency of the resulting allocation.

2009

  • Pick A Bundle: A novel bundling strategy for selling multiple items within online auctions [ext abs] [pdf]
    I. A. Vetsikas, A. C. Rogers and N. R. Jennings
    8th International Joint Conference on Autonomous Agents and Multi-agent Systems (AAMAS), Budapest, Hungary, 1225-1226.     (poster/short paper).
    In this paper, we present a novel mechanism alteration that can be used within the existing online auctions of today, that allows a seller to increase her revenue from selling a set of goods. We also show (in a forthcoming paper) that the revenue is in most cases higher than even that of using the so called "optimal auctions" (i.e. using the optimal reserve price).
    An extended version was published at OPTMAS workshop in AAMAS09. [pdf]

2008

  • Considering assymmetric opponents in multi-unit sealed-bid auctions [pdf (corrected)]
    Ioannis A. Vetsikas and Nicholas R. Jennings
    1st Market Based Control (MBC) Workshop, Liverpool, UK.
    In this paper we present ongoing work on the computation of equilibria for the case when not all bidders are the same, but rather use different utility models; more specifically we assume that the bidders have either different risk attitudes or different competitiveness (spitefulness). We also present a methodology for getting the equilibrium in this case (which we have also been working with in other auction problems). A correction to an error that has been found after publication is included here.
  • IAMwildCAT: The winning strategy for the TAC market design competition [pdf]
    Perukrishnen Vytelingum, Ioannis A. Vetsikas, Bing Shi and Nicholas R. Jennings
    18th European Conference on Artificial Intelligence (ECAI), Patras, Greece, 428-432. [Shortlisted/Finalist for best paper award]
    In this paper we examine some of the tradeoffs present in the design of a market specialist, when it competes against other specialists for traders, and we present our agent IAMwildCAT which won the first TAC CAT (market design) competition.
  • Bidding strategies for realistic multi-unit sealed-bid auctions [pdf]
    Ioannis A. Vetsikas and Nicholas R. Jennings
    23rd Conference on Artificial Intelligence (AAAI), Chicago, USA, 182-189.
    In this paper we continue our work towards analyzing auctions, when we look at a number of realistic concerns, such as budget, risk attitudes, uncertainty in the agent valuation etc. It is a continuation of [AAMAS08].
    Also published at OPTMAS workshop in AAMAS08. [pdf]
  • Towards agents participating in realistic multi-unit sealed-bid auctions [pdf]
    Ioannis A. Vetsikas and Nicholas R. Jennings
    7th International Joint Conference on Autonomous Agents and Multi-agent Systems (AAMAS), Estoril, Portugal, 1621-1624.     (poster/short paper)
    In this paper we present the initial stages of our work towards analyzing auctions, when we look at a number of realistic concerns, such as budget, risk attitudes, uncertainty in the agent valuation etc.
  • Agent-based coordination technologies in disaster management [pdf]
    Sarvapali D. Ramchurn, Alex C. Rogers, Kate Macarthur, Alessandro Farinelli, Perukrishnen Vytelingum, Ioannis A. Vetsikas and Nicholas R. Jennings
    7th International Joint Conference on Autonomous Agents and Multi-agent Systems (AAMAS), Estoril, Portugal, 1651-1652.     (demo track)
    In this demo, we presented application of agent-based technologies to the disaster management domain.

2007

  • WhiteDolphin: A TAC travel agent [pdf]
    Perukrishnen Vytelingum, Rajdeep K. Dash, Ioannis A. Vetsikas and Nicholas R. Jennings
    AAAI Workshop on Trading Agent Design and Analysis (TADA), Vancouver, Canada, 36-44.
    This paper is presents TAC agent WhiteDolphin.
  • Outperforming the competition in multi-unit sealed bid auctions [pdf]
    Ioannis A. Vetsikas and Nicholas R. Jennings
    6th International Joint Conference on Autonomous Agents and Multi-agent Systems (AAMAS), Hawaii, USA, 702-709. [Shortlisted/Finalist for best paper award]
    This paper examines what happens in multi-unit sealed-bid auctions, when agents not only wish to maximize their own profit, but also to minimize the profit of their opponents.
    Also presented at INFORMS 07, Seattle, USA (Nov 07).
  • Generating Bayes-Nash equilibria to design autonomous trading agents [pdf]
    Ioannis A. Vetsikas, Nicholas R. Jennings and Bart Selman
    20th International Joint Conference on Artificial Intelligence (IJCAI), Hyderabad, India, 1543-1550. [Nominated for best paper award]
    This paper is an extension of [EXCH06]. It provides several novel Bayes Nash equilibria and it describes the complete system of differential equations that needs to be solved in order to compute the Bayes Nash equilibria for the case of an m-th price auction that has several (8 in the case of TAC) closing times; we discover several equilibria (depending on the assumptions made by the agent). The experimental section extends the methodology in [AAMAS03].

2006

  • The ALLADIN Project: Agent Technology to the Rescue
    N. R. Jennings, S. D. Ramchurn, M. Allen-Willams, R. Dash, P. Dutta, A. C. Rogers and I. A. Vetsikas
    AAMAS Disaster Management Workshop, 157-158.
    This paper describes the work undertaken under the Aladdin project.
  • Bayes-Nash equilibria for m-th price auctions with multiple closing times [pdf]
    Ioannis A. Vetsikas and Bart Selman
    ACM SIGecom Exchanges, Vol. 6 No. 2, Dec. 2006, 27-36.
    This paper presents some novel Bayes Nash equilibria that we computed for the case of an m-th price auction that has at least 2 possible closing times.

2005

  • A principled methodology for the design of autonomous trading agents with combinatorial preferences in the presence of tradeoffs [pdf]
    Ioannis A. Vetsikas
    PhD Thesis, Cornell University, Ithaca NY, USA, August 2005.
  • Autonomous trading agent design in the presence of tradeoffs [pdf]
    Ioannis A. Vetsikas and Bart Selman
    7th International Conference on Electronic Commerce, ACM International Conference Proceeding Series Vol. 113, Xian, China, 293-299.
    In this paper we present several issues and improvements that were not presented in [AAMAS03]. We also present a complete set of experiments for both versions of the TAC game (2001-2003 and post 2004) and describe the changes necessary to the 2004 agent.

2004

  • A methodology and equilibria for design tradeoffs of autonomous trading agents [pdf]
    Ioannis A. Vetsikas and Bart Selman
    3rd International Joint Conference on Autonomous Agents and Multi-agent Systems (AAMAS), New York, USA, 1286-1287.     (short paper)
    In this paper we clarify and extend the methodology presented in [4]. One of the main differences is the introduction of strategies based on Bayes-Nash equilibria. This paper was eventually split (with a lot of additions) into papers [ICEC05], [EXCH06] and [IJCAI07].

2003

  • A principled study of the design tradeoffs for autonomous trading agents [pdf]
    Ioannis A. Vetsikas and Bart Selman
    2nd International Joint Conference on Autonomous Agents and Multi-agent Systems (AAMAS), Melbourne, Australia, 473-480.
    In this paper we present a methodology (mostly experimental) for determining the "optimal" strategy in any scenario where an agent needs to get several goods from several different auctions. There is some mention of our 2002 TAC WhiteBear entry.

2002

  • Design tradeoffs for autonomous trading agents [pdf]
    Ioannis A. Vetsikas and Bart Selman
    Proc. 8th International Conference on Principles and Practice of Constraint Programming (CP), Springer Lecture Notes in Computer Science Vol. 2470, Ithaca, NY, 790-791.     (doctoral section)
    This paper is complementary to [GTDT02] and focuses a bit more on the optimization problem that our 2001 agent faced.
  • WhiteBear: An empirical study of design tradeoffs for autonomous trading agents [pdf]
    Ioannis A. Vetsikas and Bart Selman
    AAAI Workshop on Game Theoretic and Decision Theoretic Agents (GTDT), Edmonton, Canada, 81-88.
    This paper mainly examines the problem of the TAC 2001 game and our entry into the TAC 2001 competition.

2001

  • Structure and phase transition phenomena in the VTC problem [pdf]
    Ramon Bejar, Ioannis A. Vetsikas, Carla P. Gomes, Henry Kautz and Bart Selman
    TASK-PI Darpa Workshop, Santa Fe, New Mexico, USA, April 2001.
    We examine computational issues involved in the computation of a "fair" allocation of chores to several contracted agents.

1998

  • Navigation of a vehicle in a maze using neural, fuzzy and neuro-fuzzy techniques (in Greek)
    Ioannis A. Vetsikas
    Diploma thesis, National Technical University of Athens, Athens, Greece, July 1998.

Software

I authored several versions of WhiteBear, the Cornell entry into the TAC competition.

1.     TAC Classic in 2001, 2002, 2003, 2004 and 2005 (the winner in several rounds of the competition and the finals in 2002 and 2004)

2.     TAC SCM in 2003 and 2004

I am the co-author of agent IAMwildCAT, which won the first TAC CAT competition in 2007.

Contact Info

School of Electronics and Computer Science
University of Southampton

Highfield, Southampton SO17 1BJ

UK

 

Tel (office): +44 (0) 23-8059-3256
Email : iv [at] ecs [dot] soton [dot] ac [dot] uk