Bernardo A. Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....we can think of those programs as the multiple restarts of the same algorithm. However, unlike restart, programs in a cooperative search interact with each other based on a cooperation protocol that speci es how the search programs cooperate at run time. Intuitively, as framed in Huberman s paper [10], cooperation is an exchange of hints that may confuse some search processes, but will also help others. Overall, hint sharing improves the performance and has been used with success to design search heuristics in the context of constraint satisfaction problems [3] 8] and to parallelize some ....

....cant compared to the time required by the search heuristics. B. Convergence behavior of CoMHP Because of the local interactions among the search processes, the convergence behavior of cooperative algorithms is sometime modeled according to the theory of complex systems. For example, Huberman [10] uses a probabilistic model to show that the performance of cooperating processes is log normally distributed for successful cooperative algorithms, in contrast with the normal distribution of independent searches (or restarts) In this model, the e ect of cooperation on the distribution is a ....

B.A. Huberman. The Performance of Cooperative Processes. Physica D, 42:38-47, 1990.

....as well as other methods. For a particularly important category of such parallel algorithms, data gathered by a sequential search program is recycled and shared with other concurrently executing search programs. In the literature, these parallel algorithms are referred to as cooperative search [3,4,19,20], multiple interacting walks [40] or population approaches [28] Procedures such as coarse grained parallel genetic algorithms (PGA) 17,33,35] and ne grained PGA [16,26,30] Memetic Algorithms [6,27,29] parallel Branchand Bound [13] tabu search algorithms [8,10,11,23] simulating annealing ....

....solutions. Similar results have been obtained in many other elds [7] However, while researchers acknowledge that cooperation has an impact on the search pattern of cooperating search processes, only limited e orts to develop a comprehensive model of cooperation among search processes (Huberman [20], Clearwater all [3] have been undertaken so far. This is re ected, for example, in the way cooperative algorithms are designed, often simply by adding communication extensions to existing sequential search programs. We believe these ad hoc approaches fail to identify important design ....

B.A. Huberman. The Performance of Cooperative Processes. Physica D, 42:38{ 47, 1990.

....strategy as a part of new semantics of the assignment is invoked automatically. Deciding communication structure and sending information is done in meta level. 15 4.3. related work Huberman stated that cooperation between searching agents improves performance through theoretical analysis[12, 13]. Hogg also researched cooperation in a graph problem[11] But they did not mention how to communicate with each other and did not consider communication costs. Their works are a kind of cooperation in parallel activities. Need for cooperation is not emerged from the constraint of execution ....

Bernardo A. Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990. Reprinted in [7].

....used in artificial life models, is to strip away as much detail as possible, retaining only the essential interactions. This article describes one such model, called Echo, whose emphasis is on the essential components and interactions of ecosystems. Many CAS can be viewed as ecologies (e.g. [24, 46]) but our focus in this article is on the analogy with natural ecologies. Our goal for Echo is to capture in an artificial ecosystem a set of behaviors that are robust with respect to the details of the interactions (e.g. avoiding parameter tweaking to coax a system to produce desired behaviors) ....

Huberman, B. A. (1991). The performance of cooperative processes. In S. Forrest (Ed.), Emergent computation (pp. 38--47). Cambridge, MA: MIT Press.

....all agents are allowed to discount based on utility, they have access to the cost of the task to all other agents. This is similar to a system of agents that give hints to each other. It has been shown that the exchange of hints among agents can allow some agents to get more resources than others [14]. How important this unfairness becomes depends on the full strategies of the agents. For our simulation with discount based on utility, see gure 5, the systems stays fair. 3 0 100 200 300 400 500 600 700 800 900 1000 0 1000002000003000004000005000006000007000008000009000001e 006 ....

B. A. Huberman. The performance of cooperative processes. Physica D, 42(1-3):38-47, 1990.

....of failure p(t) to find the optimal solution before iteration t [1, 15] As reported in the taxonomy of Crainic, Toulouse and Gendreau [5] the parallelization of a method such as TS can also be realized based on knowledge sharing and cooperation between the different search threads. Huberman [10] conjectures 1 a performance improvement when concurrent search threads are engaged in a cooperative strategy rather than performing independent searches. Experimental observations with concurrent asynchronous interactive TS [4] and simulated annealing procedures [9] seem to confirm the ....

B.A. Huberman. The Performance of Cooperative Processes. Physica D, 42:38--47, 1990.

....we must remark on the work done at the Dynamics of Computation Group at Xerox PARC, by B. Huberman and coworkers who have shown in a series of papers some interesting results that may explain the observed features of our studies in combinatorial optimization settings [70] 71] 66] 77] 72] [73]. They found that a population of cooperative computational processes has a universal distribution of performance which can be increased by crossing sharp phase transitions in the topological structure of the problem. The relevance of natural hierarchies viewed as hints that the optimizing ....

B.A. Huberman, The Performance of Cooperative Processes, Physica D 42 (1990) 38-47.

....discuss the complexity of automatically deriving social laws for agent groups. They show that the problem is NP complete but can, under a number of restrictions, be made polynomial. Some DAI work draws heavily from mathematical results in the field of parallel distributed systems. In particular, Huberman (1990) describes the effects of information exchange on the performance of a collection of agents applied to a class of search problems. He also addresses the ubiquity of log normal distributions of performance found across different domains, and hypothesizes a universal law of distribution for all ....

....field that studies computational ecosystems, simulations of populations of agents with well defined interactions. The research is focused on global effects and the changes in the system as a whole over time. This process of global changes is usually referred to co evolution (Kephart, Hogg Huberman 1990). Often the systems studied have some similarities to the global effects found in biological ecosystems, but the complex details of biological systems cannot be reasonably addressed. Co evolution experiments are used to find improved search based optimization techniques. For example, Hillis ....

Huberman, B. A. (1990), `The Performance of Cooperative Processes', Physica D 42, 38--47.

....can be transparent and powerful. Finally, our particular result regarding the importance of population diversity for efficiency of multi agent systems has been a theme in biologically inspired computer system design. Huberman has studied performance characteristics of cooperative processes [8], looking at the concurrent search of a space of possibilities by an ecology of agents and at the effectiveness of hints exchanged among those agents. He also concludes that diversity increases effectiveness. In the field of genetic algorithms [11] the tradeoff between population diversity and ....

B. A. Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990.

....to make it easier to integrate parallel computation and traditional economic models. See Mount and Reiter (1990, 1994, 1996) Computer scientists have also seen the natural interpretation of networks of machines as networks of boundedly rational human agents (e.g. Huberman and Hogg (1988) and Huberman (1990)) Subsequent research by economists includes the use of models of decentralized associative computation to study the structure of organizations. The first example was Radner (1993) followed by Bolton and Dewatripont (1994) Van Zandt (1998b) Meagher (1996a) Prat (1997) and Orbay (1997) among ....

Huberman, B. A. (1990). The performance of cooperative processes. Physica, 42, 38--47.

....high level and the low level goals of the system. We experimentally demonstrate that our boxpushing scenarios falls into the class of problems in which communication can produce superlinear performance improvements on problems that otherwise allow only linear speed up with cooperative solutions (Huberman 1990). We give strong evidence that single robot and non communicating two robot solutions are inferior, in terms of performance, compared to the communication and turn taking protocol given in this paper. 2 Experimental Environment We used two commercially available Genghis II sixlegged robots in ....

Huberman, B. A. (1990), `The Performance of Cooperative Processes', Physica D 42, 38--47.

.... from those we studied in that they are neither embodied nor situated in a simulated or natural physical world (for an overview see Gasser Huhns (1989) Other branches of DAI deal with simpler distributed systems, focusing on the role of cooperation and competition in the multi agent environment (Huberman 1990). In robotics, the last decade has witnessed a shift in the emphasis of research away from purely theoretical and simulated work toward physical implementations akin to ours. Most of the work in robotics is focused on control of a single agent, but several groups have obtained and experimented ....

Huberman, B. A. (1990), `The Performance of Cooperative Processes', Physica D 42, 38--47.

....used in artificial life models, is to strip away as much detail as possible, retaining only the essential interactions. This paper describes one such model, called Echo, whose emphasis is on the essential components and interactions of ecosystems. Many CAS can be viewed as ecologies (e.g. [24, 46]) but our focus in this paper is on the analogy with natural ecologies. Our goal for Echo is to capture in an artificial ecosystem a set of behaviors that are robust with respect to the details of the interactions (e.g. avoiding parameter tweaking to coax a system to produce desired behaviors) ....

B. A. Huberman. The performance of cooperative processes. In S. Forrest, editor, Emergent Computation, pages 38--47, Cambridge, MA, 1991. MIT Press.

....Echo was designed to capture the essential features of ecological systems in an agentbased model. All of the entities and interactions in Echo are highly abstract, and it is not yet known whether Echo can be used to model real world phenomena effectively. Many CAS can be viewed as ecologies (e.g. [8]) but our focus in this paper is on the analogy with natural ecologies. Echo resembles some other CAS models. These include Swarm [9] Sugarscape [10] and the Evolutionary Reinforcement Learning (ERL) model [11] Unlike Swarm, Echo makes specific commitments about agent types and interactions; ....

B.A. Huberman. The performance of cooperative processes. In S. Forrest, editor, Emergent Computation, pages 38--47, Cambridge, MA, 1991. MIT Press.

....dynamics of the robot and box interaction are de3 signed so as to increase efficiency in terms of forward movement of the box. We will demonstrate a communication and turn taking protocol, described in a later section, that satisfies both the high level and the low level goals of the system. Huberman (1990) shows that the introduction of communication can produce superlinear performance improvements on problems that otherwise allow only linear efficiency improvements through the use of cooperative solutions. We will experimentally demonstrate that our box pushing scenario falls into that class of ....

Huberman, B. A. (1990), `The Performance of Cooperative Processes', Physica D 42, 38--47.

....rational agents. Certain aspects of DAI work are purely theoretical and address the difficulty of multi agent planning and control in abstract environments (e.g. Shoham Tennenholtz (1992) Some DAI work draws heavily from mathematical results in the field of parallel distributed systems (e.g. Huberman (1990), Clearwater, Huberman Hogg (1991) and many others) DAI and Alife merge in the experimental mathematics field that studies computational ecosystems, using simulations of populations of agents with well defined interactions. The research is focused on global effects and the changes in the ....

....that studies computational ecosystems, using simulations of populations of agents with well defined interactions. The research is focused on global effects and the changes in the system as a whole over time. This process of global changes is usually referred to as co evolution (Kephart, Hogg Huberman 1990). Co evolution experiments are usually used to find improved search based optimization techniques (Hillis 1990) Often the systems studied have some similarities to the global effects found in biological ecosystems, but the complex details of biological systems cannot be reasonably addressed. ....

Huberman, B. A. (1990), `The Performance of Cooperative Processes', Physica D 42, 38--47.

....but the choices made here illustrate the potential of this method and have many direct correspondences with a wide range of constraint satisfaction problems. As a note of caution in developing more sophisticated strategies, the choices made should tend to promote high diversity among the agents [7, 9] so there will be many opportunities to try hints in di#erent promising contexts. This means that some choices that appear reasonable when viewed from the perspective of a single agent, could result in lowered performance for the group as a whole. 2.2.1 cooperative search There are two basic ....

Bernardo A. Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990.

....problem solving heuristic can be introduced into a different problem solving heuristic. A general theory predicts superlinear speedup in the performance of individual agents when the search methods are diverse and the agents are able to utilize information found in other parts of the search space [28]. The current problem of portfolio selection is interesting because it is a stochastic online decision problem [30] agents must invest as they receive incremental information about stock prices. Many other problem solving techniques that are related to cooperative search have been proposed for ....

Bernardo A Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990. 40 PARKES & HUBERMAN

No context found.

Bernardo A. Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990.

....information that can be introduced into the problem solving heuristic of another agent. A general theory predicts superlinear speedup in the performance of individual agents when the search methods are diverse and the agents are able to utilize information found in other parts of the search space (Huberman, 1990). In our domain of multiagent portfolio selection in an unknown environment all agents receive the same incremental information about the market in each investment period and have limited computational resources with which to process the information. The flow of new information creates a ....

Huberman, B. A. (1990). "The performance of cooperative processes," Physica D 42, 38--47.

....of individual portfolio selection strategies. An agent can switch to the portfolio strategy of the agent that has been performing best in the recent past. This simple mechanism of hint exchange has enabled exponential performance improvements in other cooperative problem solving domains (Huberman 1990; Clearwater, Huberman, Hogg 1991) We present the results of a quantitative assessment of the performance of our multiagent portfolio selection model in a simple stochastic market that show that: a) a system of diverse agents will outperform a single agent; b) a system of agents can further ....

.... 1993; Aldous Vazirani 1994; Hogg Williams 1993; Clearwater, Huberman, Hogg 1991) A general theory predicts superlinear speedup in the performance of individual agents when the search methods are diverse and the agents are able to utilize information found in other parts of the search space (Huberman 1990). 5 0 5 10 15 20 25 0 20 40 60 80 100 Log(Final Wealth) Frequency Market Portfolio Adaptive Agents Figure 17: Distribution of final log wealth of the market portfolio and a system of 400 adaptive agents in a simulated CAPM market. Schaerf et.al. study communication within a multiagent system ....

Huberman, B. A. 1990. The performance of cooperative processes. Physica D 42:38--47.

No context found.

Bernardo A. Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990.

No context found.

Bernardo A. Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990.

No context found.

B. A. Huberman. The performance of cooperative processes. Physica D, 42:38--47, 1990.

No context found.

Bernardo A. Huberman (1990). The Performance of Cooperative Processes. Physica D 42, 38-37.

First 50 documents

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