Juill H., Pollack J.B.: Massively Parallel Genetic Programming. In: Angeline P.J., Kinnear K.E. (eds): Advances in Genetic Programming 2, Chapter 17. MIT Press, Cambridge, MA (1996) 339-358  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to compile at runtime standard GP trees into machine code before evaluation. Speed up strategies based on intelligently reducing the number of fitness cases have been proposed [5, 24, 11] Finally, some research has been devoted to parallel and distributed implementations of GP (see for example [1, 23, 7]) Some of these techniques are now used in many GP implementations. This and the increased power of modern workstations make it possible run 50 generations of a typical GP benchmark problem with a population of 500 individuals in perhaps ten seconds on a normal workstation. Nonetheless, the ....

Hugues Juille and Jordan B. Pollack. Massively parallel genetic programming. In Peter J. Angeline and K. E. Kinnear, Jr., editors, Advances in Genetic Programming 2, chapter 17, pages 339-358. MIT Press, Cambridge, MA, USA, 1996.

.... produced by ADFs the first time they are run with a certain set of arguments, and using the stored results thereafter [Langdon, 1998] Finally, some research has been devoted to parallel and distributed implementations of GP (see for example [Andre and Koza, 1996; Stoffel and Spector, 1996; Juille and Pollack, 1996; Sian, 1998] These are usually based on the idea of distributing a population across multiple machines with some form of communication between them to exchange useful genetic material. A similar, but more basic, speed up technique is to perform independent multiple runs of a same problem on ....

Juille, H. and Pollack, J. B. (1996), "Massively parallel genetic programming," in Advances in Genetic Programming 2, P. J. Angeline and K. E. Kinnear, Jr. (Eds.), Chapter 17, pp 339--358, Cambridge, MA, USA: MIT Press.

.... on intelligently reducing the number of fitness cases have been proposed (Gathercole and Ross, 1997; Teller and Andre, 1997; Langdon, 1998) Finally, some research has been devoted to parallel and distributed implementations of GP (see for example (Andre and Koza, 1996; Stoffel and Spector, 1996; Juille and Pollack, 1996)) Some of these techniques are now used in many GP implementations. This and the increased power of modern workstations make it possible run 50 generations of a typical GP benchmark problem with a population of 500 individuals in perhaps ten seconds on a normal workstation. Nonetheless, the ....

Juille, H. and Pollack, J. B. (1996). Massively parallel genetic programming. In Angeline, P. J. and Kinnear, Jr., K. E., editors, Advances in Genetic Programming 2, chapter 17, pages 339--358. MIT Press, Cambridge, MA, USA.

....to compile at runtime standard GP trees into machine code before evaluation. Speed up strategies based on intelligently reducing the number of fitness cases have been proposed [5, 24, 11] Finally, some research has been devoted to parallel and distributed implementations of GP (see for example [1, 23, 7]) Some of these techniques are now used in many GP implementations. This and the increased power of modern workstations make it possible run 50 generations of a typical GP benchmark problem with a population of 500 individuals in perhaps ten seconds on a normal workstation. Nonetheless, the ....

Hugues Juille and Jordan B. Pollack. Massively parallel genetic programming. In Peter J. Angeline and K. E. Kinnear, Jr., editors, Advances in Genetic Programming 2, chapter 17, pages 339--358. MIT Press, Cambridge, MA, USA, 1996.

....(e.g. wall following) and as a form of concept learning on the symbolic level (e.g. genetic logic programming, Wong, 21] It will be assumed here that massively parallel on line evolution for realistic problems is a future possibility with GP. For current work on this see e.g. Juiles et al. [11] for massively parallel GP, and Steels [20] for on line evolution. It will also be assumed here that data structures can be evolved (a detailed report is given in Langdon [14] The translations f A and f C can be implemented using GP. When the fitness of a single genetic program (which we ....

Juille,H., Pollack,J.B. (1996) "Massively Parallel Genetic Programming" in Advances in Genetic Programming 2, edited by Peter J. Angeline and Kenneth E. Kinnear , MIT Press.

....Hardware In common with GAs, the vast proportion of machine resources consumed by GP are used running the fitness function to evaluate the population. Like GAs, GP parallelises easily and can readily take advantage of a wide range of dedicated parallel processing architectures [TOD96, SS96, JP96, Ikr96, AK96, OCPT96] There is also interest in parallel execution using networks of workstations (e.g. using PVM) and across the Internet using Java. Often GP (and GAs) are parallelised by splitting the population so different parts of it are run on different CPUs. This also has the potential ....

Hugues Juille and Jordan B. Pollack. Massively parallel genetic programming. In Peter J. Angeline and K. E. Kinnear, Jr., editors, Advances in Genetic Programming 2, chapter 17, pages 339--358. MIT Press, Cambridge, MA, USA, 1996.

.... in PDGP as graphs with nodes representing functions and terminals and links representing 1 The development of parallel programs should not be confused with the parallel implementations of GP, which are essentially methods of speeding up the genetic search of standard tree like programs [3, 6, 21, 13, 25]. These methods are usually based on the use of multiple processors, each one handling a separate population, a subset of fitness evaluations or a subset of fitness cases. max 3 x y Figure 2: Graph like representation of the expression max(x y, 3 x y) the flow of control and ....

Hugues Juille and Jordan B. Pollack. Massively parallel genetic programming. In Peter J. Angeline and K. E. Kinnear, Jr., editors, Advances in Genetic Programming 2, chapter 17. MIT Press, Cambridge, MA, USA, 1996.

.... produced by ADFs the first time they are run with a certain set of arguments, and using the stored results thereafter (Langdon, 1998) Finally, some research has been devoted to parallel and distributed implementations of GP (see for example (Andre and Koza, 1996; Stoffel and Spector, 1996; Juille and Pollack, 1996)) These are usually based on the idea of distributing a population across multiple machines with some form of communication between them to exchange useful genetic material. A similar, but more basic, speed up technique is to perform independent multiple runs of a same problem on different ....

Juille, H. and Pollack, J. B. (1996). Massively parallel genetic programming. In Angeline, P. J. and Kinnear, Jr., K. E., editors, Advances in Genetic Programming 2, chapter 17, pages 339--358. MIT Press, Cambridge, MA, USA.

No context found.

Juille, H. and Pollack, J. (1995). Massively parallel genetic programming. In Angeline, P. and Kinnear, K., editors, Advances in Genetic Programming II. MIT Press, cambridge.

....and recombination, using the kernel of the parallel GP described in the previous section. In this paper, only a tournament style of competitive evolution has been used and compared to canonical GP. A more general presentation of the different strategies that have been implemented can be found in [Juill e Pollack, 1996]. 3 The Spiral Problem and the Competitive Evolution Paradigm 3.1 Presentation The intertwined spiral problem consists of learning to classify points on the plane into two classes according to two intertwined spirals. The data set is composed of two sets of 97 points, on the plane between 7 and ....

....summing all their scores in the competition. Once individual fitness is evaluated, selection and recombination are performed according to a fitness proportionate rule. Details of the implementation of this model of tournament, and of selection and recombination procedures for MPGP can be found in [Juill e Pollack, 1996]. 3.2 Preliminary Results and Discussion For the two classes of experiments, we performed 25 runs and each run was stopped after 300 generations. At each generation, 90 of the population was replaced by offsprings resulting from recombination and the remaining 10 was the result of fitness ....

Juill'e, H. & Pollack, J. B. (1996). Massively parallel genetic programming. In Angeline & Kinnear (Eds.), Advances in Genetic Programming II. MIT Press.

.... was explored by Hillis (Hillis, 1992) on the sorting problem, by Angeline Pollack (Angeline and Pollack, 1994) on genetically programmed tic tac toe players, on predator prey games, e.g. Cliff and Miller, 1995, Reynolds, 1994) and by Juille Pollack on the intertwined spirals problem (Juille and Pollack, 1995). Rosin Belew applied competitive fitness to several games (Rosin and Belew, 1995) However, besides Tesauro s TD Gammon, which has not to date been viewed as an instance of co evolutionary learning, Sims artificial robot game (Sims, 1994) is the only other domain as complex as Backgammon to ....

Juille, H. and Pollack, J. (1995). Massively parallel genetic programming. In Angeline, P.

....and recombination, using the kernel of the parallel GP described in the previous section. In this paper, only a tournament style of competitive evolution has been used and compared to canonical GP. A more general presentation of the different strategies that have been implemented can be found in [Juill e Pollack, 1995]. 3 The Spiral Problem and the Competitive Evolution Paradigm Experiments were conducted to compare canonical GP evolution to competitive evolution for the intertwined spiral problem. This learning problem consists in classifying points into two classes according to two intertwined spirals. The ....

....individuals are very likely to be different from one tournament to the other. Once individual fitness is evaluated, selection and recombination are performed according to a fitness proportionate rule. Details of the implementation of selection and recombination procedures for MPGP can be found in [Juill e Pollack, 1995]. If (4 x 2 Gamma y 2 ) 0:0 then return (sin( Gamma3:0 y) else return sin( 0:3214x 0:04762 Gammacos(sin( y x 0:7874) endif Figure 4: Interpretation of the solution for the intertwined spiral problem. 4 Results and Discussion For the two classes of experiments, we ....

Juill'e, H. & Pollack, J. B. (1995). Massively parallel genetic programming. In Angeline & Kinnear (Eds.), Advances in Genetic Programming II. MIT Press. To appear.

.... the sorting network problem (Hillis, 1992) on tic tac toe and other strategy games (Angeline and Pollack, 1994, Rosin and Belew, 1995, Schraudolph et al. 1994) on predator prey games (Cliff and Miller, 1995, Reynolds, 1994) and on classification problems such as the intertwined spirals problem (Juille and Pollack, 1995). However, besides Tesauro s TD Gammon, which has not to date been viewed as an instance of co evolutionary learning, Sims artificial robot game (Sims, 1994) is the only other domain as complex as backgammon to have had substantial success. Since a weak player can sometimes defeat a strong one, ....

Juille, H. and Pollack, J. (1995). Massively parallel genetic programming. In Angeline, P. and Kinnear, K., editors, Advances in Genetic Programming II. MIT Press, cambridge.

.... the sorting network problem (Hillis, 1992) on tic tac toe and other strategy games (Angeline and Pollack, 1994, Rosin and Belew, 1995, Schraudolph et al. 1994) on predator prey games (Cliff and Miller, 1995, Reynolds, 1994) and on classification problems such as the intertwined spirals problem (Juille and Pollack, 1995). However, besides Tesauro s TD Gammon, which has not to date been viewed as an instance of co evolutionary learning, Sims artificial robot game (Sims, 1994) is the only other domain as complex as backgammon to have had substantial success. Since a weak player can sometimes defeat a strong one, ....

Juille, H. and Pollack, J. (1995). Massively parallel genetic programming. In Angeline, P. and Kinnear, K., editors, Advances in Genetic Programming II. MIT Press, cambridge.

.... This was explored by Hillis (Hillis, 1992) on the sorting problem, by Angeline Pollack (Angeline and Pollack, 1994) on genetically programmed Tic Tac Toe players, on predator prey games, e.g. Cliff and Miller, 1995, Reynolds, 1994) and by Juille Pollack on the intertwined spirals problem (Juille and Pollack, 1995). Rosin Belew applied competitive fitness to several games (Rosin and Belew, 1995) However, besides Tesauro s TDGammon, which has not to date been viewed as an instance of co evolutionary learning, Sims artificial robot game (Sims, 1994) is the only other domain as complex as Backgammon to ....

Juille, H. and Pollack, J. (1995). Massively parallel genetic programming. In Kinnear, P. A. . K., editor, Advances in Genetic Programming II. MIT Press, cambridge.

No context found.

Juill H., Pollack J.B.: Massively Parallel Genetic Programming. In: Angeline P.J., Kinnear K.E. (eds): Advances in Genetic Programming 2, Chapter 17. MIT Press, Cambridge, MA (1996) 339-358

No context found.

Juille, H. and Pollack, J. (1996). Massively parallel genetic programming. In Angeline, P. and Kinnear, Jr., K., editors, Advances in Genetic Programming 2, pages 339--358. The MIT Press, Cambridge, MA, USA.

No context found.

Hugues Juille and Jordan B. Pollack. Massively parallel genetic programming. In Peter J. Angeline and K. E. Kinnear, Jr., editors, Advances in Genetic Programming 2, chapter 17. MIT Press, Cambridge, MA, USA, 1996.

No context found.

Hugues Juille and Jordan B. Pollack. Massively parallel genetic programming. In Peter J. Angeline and K. E. Kinnear, Jr., editors, Advances in Genetic Programming 2, chapter 17. MIT Press, Cambridge, MA, USA, 1996.

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