Michael P. Wellman and Junling Hu. Conjectural equilibrium in multiagent learning. Machine Learning, 33(2-3):179--200, 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....from buyers. An improvement was suggested, that the history should be large, but more weight placed on recent events [Kephart et al. A recursive algorithm can be described as an agent building up a model about what it thinks about other agents, and what those agents think about further agents [Hu et al.] This is unfeasible with a large number of evolving participants as the number of models built up increases exponentially and would become unmaintainable in my opinion. One final idea is to use an agent that is based on a mathematical model, but is adaptive in that it can decide not to use that ....

J. Hu and M. Wellman, "Conjectural Equilibrium in Multi Agent Learning", published 1998.

....Fictitious Play [20] and other models of myopic best response dynamics [57, 33] where agents play a best response to a model that they learn of their opponents. Recent models of multiagent learning within artificial intelligence provide a hierarchy of agent models and allow strategic learning [21, 56, 55], where agents take advantage of models of the learning of other agents. Instead, we follow the framework of model free learning: the agents in our model do not maintain an explicit model of the stock market. The current portfolio of an agent represents the cumulative learning of the agent, and ....

Michael P Wellman and Junling Hu. Conjectural equilibrium in multiagent learning. Machine Learning, 33:179--200, 1998.

.... the question of how much modeling of other producers is enough, or, alternatively, when does further modeling of other producers fail to help, or even harm a producer s progress (Vidal Durfee 1998a) provides hope that a producer may be able to shallowly model another producer and pro t, while (Wellman Hu 1998) warns that agents which try to leanr models of each other could wind up being worse o than if they had not learned at all. Clearly, this is a situation to be avoided if possible. The previous example also assumed that the producer and consumer shared the same taxonomy of articles. This seems ....

Wellman, M., and Hu, J. 1998. Conjectural equilibrium in multiagent learning. Machine Learning 33.

.... the question of how much modeling of other producers is enough, or, alternatively, when does further modeling of other producers fail to help, or even harm a producer s progress (Vidal Durfee 1998a) provides hope that a producer may be able to shallowly model another producer and profit, while (Wellman Hu 1998) warns that agents which try to learn models of each other could wind up being worse off than if they had not learned at all. Clearly, this is a situation to be avoided if possible. The previous example also assumed that the producer and consumer shared the same taxonomy of articles. This seems ....

Wellman, M., and Hu, J. 1998. Conjectural equilibrium in multiagent learning. Machine Learning 33.

....auctions with entry and exit of agents, as ocurrs in the UMDL. To address the question of what an agent should do in multiagent interactions, some researchers have developed a new solution concept based on a recursive model of other agents (i.e. what they think about what I think about and so on) [2, 3, 9]. While gametheoretic agents are ultra smart and super rational such that they can reason about this whole recursive hierarchy ad infinitum [10] the recursive modeling method (RMM) assumes that agents can only build a finite nesting of models (because of practical limitations on acquiring such ....

J. Hu and M. P. Wellman, "Conjectural Equilibrium in Multiagent Learning," Machine Learning, vol. 33, 1998.

....unrealistic for many complex systems. Instead, by studying how the individual, strategic agents impact the overall system behavior, we gain insights on properties of agent societies, such as characterizing the types of environments and agent populations that foster social and anti social behavior [25, 26]. Our approach falls into this category, as one of our goals is to explain how strategic agents (in particular the p strategy sellers) affect the UMDL in terms of market and allocation efficiency. Studies on system level analyses of the multiple agents, investigating the roles of system wide ....

....as one of our goals is to explain how strategic agents (in particular the p strategy sellers) affect the UMDL in terms of market and allocation efficiency. Studies on system level analyses of the multiple agents, investigating the roles of system wide properties are found in other papers [9, 10, 12, 13, 26]. Some results show that in large markets with significant uncertainty, taking ones effect on prices into account does not pay off. This is, of course, a positive indication, since it implies that agents in large markets are not encouraged to speculate about their effect on prices. At the same ....

Wellman, M.P. and J. Hu, Conjectural Equilibrium in Multiagent Learning. Machine Learning, 1998. 33: 179-200.

.... studies in multi agent learning focus on settings where payo s are either negatively correlated, as in zero sum games (see, for example, Littman [33] or positively correlated, as in coordination games (see, for example, Shoham and Tennenholtz [45] Two notable exceptions include Wellman and Hu [50], who conducted a theoretical investigation of multi agent learning in market interactions, and Sandholm and Crites [42] who conducted empirical studies of multi agent reinforcement learning in the Prisoners Dilemma. Similarly, this work considers multi agent learning in positive sum games. The ....

M. Wellman and J. Hu. Conjectural equilibrium in multi-agent learning. Machine Learning, 33:179-200, 1998. Special Issue on Multi-agent Learning. 58

.... (Good, 1971; Simon, 1976; Boddy and Dean, 1989; Russell and Wefald, 1991) Recent models of multiagent learning provide a hierarchy of agent models and allow an analysis of the effect of strategic (non myopic) learning on the equilibrium outcomes in games (Gmytrasiewicz and Durfee, 1995; Wellman and Hu, 1998; Vidal and Durfee, 1998) We model bounded rational agents, that cannot compute optimal investment strategies directly and are able to select good strategies more effectively with cooperation from other bounded rational agents. The problem that faces an agent in our portfolio selection problem ....

Wellman, M. P., and Hu, J. (1998). "Conjectural equilibrium in multiagent learning," Machine Learning 33, 179--200.

....with subtask decomposition with resource dependencies. This model includes agent capabilities, preferences, costs, properties of solutions, and a global optimality metric efficiency for evaluating solution 8 quality. This chapter directly includes material presented in [Walsh and Wellman, 1998] and further developed in [Walsh and Wellman, 1999; Walsh et al. 2000] In Chapter 3, I describe price systems as an abstract framework for guiding decentralized agent decision making by encapsulating information about the relative value of goods. Competitive price equilibria are well known ....

....a major shortcoming of SAMP SB is that dead ends can form, in which agents acquire inputs but are unable to sell their outputs. I propose an augmented protocol, SAMP SB D, that allows agents to decommit from the undesirable input contracts. This chapter directly includes material from [Walsh and Wellman, 1998; 1999] 9 In Chapter 5, I describe an alternate approach to address the bid miscoordination problem inherent in distributed auctions. I present a particular combinatorial auction that, given indivisible bids from agents specifying their entire demand for all goods, computes allocations over ....

[Article contains additional citation context not shown here]

Michael P. Wellman and Junling Hu. Conjectural equilibrium in multiagent learning. Machine Learning, 33:179--200, 1998.

....we randomly chose producer costs uniformly over [0; 1] and for each producer, based its bid on 500 Monte Carlo samples over the costs of other producers in its instance. We also 5 It is well established that pursuing aggressively strategic behavior can lead to pitfalls in uncertain environments [12, 16]. Proposers of particular bidding behaviors (strategic or not) in particular environments still face the burden of demonstrating reasonableness. a1 a2 a3 a9 a10 1 3 9 2 a4 a5 a6 a7 a8 4 5 6 7 8 a11 a12 a13 a14 a15 a16 1 0 1 1 1 2 a17 a18 1 3 c1 a19 a20 1 4 a21 ....

M. P. Wellman and J. Hu. Conjectural equilibrium in multiagent learning. Machine Learning, 33:179--200, 1998.

No context found.

Michael P. Wellman and Junling Hu. Conjectural equilibrium in multiagent learning. Machine Learning, 33(2-3):179--200, 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC