%0 Journal Article %1 Salustowicz:97ecj %A Salustowicz, R. P. %A Schmidhuber, J. %D 1997 %J Evolutionary Computation %K Evolution, Incremental Learning, Population-Based Program algorithms, environments genetic languages, observable partially probabilistic program programming programming, search, stochastic %N 2 %P 123--141 %R doi:10.1162/evco.1997.5.2.123 %T Probabilistic Incremental Program Evolution %U ftp://ftp.idsia.ch/pub/rafal/PIPE.ps.gz %V 5 %X Probabilistic Incremental Program Evolution (PIPE) is a novel technique for automatic program synthesis. We combine probability vector coding of program instructions, Population-Based Incremental Learning, and tree-coded programs like those used in some variants of Genetic Programming (GP). PIPE iteratively generates successive populations of functional programs according to an adaptive probability distribution over all possible programs. Each iteration it uses the best program to refine the distribution. Thus, it stochastically generates better and better programs. Since distribution refinements depend only on the best program of the current population, PIPE can evaluate program populations efficiently when the goal is to discover a program with minimal runtime. We compare PIPE to GP on a function regression problem and the 6-bit parity problem. We also use PIPE to solve tasks in partially observable mazes, where the best programs have minimal runtime.

@article{Salustowicz:97ecj, abstract = {Probabilistic Incremental Program Evolution (PIPE) is a novel technique for automatic program synthesis. We combine probability vector coding of program instructions, Population-Based Incremental Learning, and tree-coded programs like those used in some variants of Genetic Programming (GP). PIPE iteratively generates successive populations of functional programs according to an adaptive probability distribution over all possible programs. Each iteration it uses the best program to refine the distribution. Thus, it stochastically generates better and better programs. Since distribution refinements depend only on the best program of the current population, PIPE can evaluate program populations efficiently when the goal is to discover a program with minimal runtime. We compare PIPE to GP on a function regression problem and the 6-bit parity problem. We also use PIPE to solve tasks in partially observable mazes, where the best programs have minimal runtime.}, added-at = {2008-06-19T17:46:40.000+0200}, author = {Salustowicz, R. P. and Schmidhuber, J.}, biburl = {https://www.bibsonomy.org/bibtex/2c36088d95f5a53dc0c6dbaa2ed77b1b3/brazovayeye}, doi = {doi:10.1162/evco.1997.5.2.123}, interhash = {f90e46778b7643a5390cafe62e85245d}, intrahash = {c36088d95f5a53dc0c6dbaa2ed77b1b3}, journal = {Evolutionary Computation}, keywords = {Evolution, Incremental Learning, Population-Based Program algorithms, environments genetic languages, observable partially probabilistic program programming programming, search, stochastic}, notes = {Date: Tue, 20 Jan 1998 19:49:23 +0100 (MET) PIPE_v1.0 SOFTWARE PACKAGE Rafal Salustowicz IDSIA, Switzerland Probabilistic Incremental Program Evolution (PIPE) is a novel technique for automatic program synthesis. Details are described in R.P. Salustowicz and J. Schmidhuber, Probabilistic Incremental Program Evolution, Evolutionary Computation, 5(2):123-141, 1997. See ftp://ftp.idsia.ch/pub/rafal/PIPE.ps.gz The PIPE_v1.0 software package allows for applying and modifying the PIPE paradigm, and for benchmarking. URLs are: http://www.idsia.ch/~rafal/research.html ftp://ftp.idsia.ch/pub/rafal/PIPE_v1.0.tar.gz The software is written in C. It - is easy to install (comes with an automatic installation tool). - is easy to use: setting up PIPE_V1.0 for different problems requires a minimal amount of programming. User-written, application- independent program parts can easily be reused. - is efficient: PIPE_V1.0 has been tuned to speed up performance. - is portable: comes with source code (optimized for SunOS 5.5.1). - is extensively documented(!) and contains three example applications. - supports statistical evaluations: it facilitates running multiple experiments and collecting results in output files. - includes testing tool for testing generalization of evolved programs. - supports floating point and integer arithmetic. - has extensive output features. - For lil-gp users: Problems set up for lil-gp 1.0 can be easily ported to PIPE_v1.0. The testing tool can also be used to process programs evolved by lil-gp 1.0. PIPE_v1.0 is licensed free of charge for non-profit or internal use. ********************************************************************** * Rafal Salustowicz * * Istituto Dalle Molle di Studi sull'Intelligenza Artificiale (IDSIA)* * Corso Elvezia 36 e-mail: rafal@idsia.ch * * 6900 Lugano ============== * * Switzerland raf@cs.tu-berlin.de * * Tel (IDSIA) : +41 91 91198-38 raf@psych.stanford.edu * * Tel (office): +41 91 91198-32 * * Fax : +41 91 91198-39 WWW: http://www.idsia.ch/~rafal * **********************************************************************}, number = 2, pages = {123--141}, size = {17 pages}, timestamp = {2008-06-19T17:50:57.000+0200}, title = {Probabilistic Incremental Program Evolution}, url = {ftp://ftp.idsia.ch/pub/rafal/PIPE.ps.gz}, volume = 5, year = 1997 }