Optimal Ordered Problem Solver

Created by W.Langdon from gp-bibliography.bib Revision:1.3872

@TechReport{schmidhuber:2002:TR12,
  author =       "Juergen Schmidhuber",
  title =        "Optimal Ordered Problem Solver",
  institution =  "IDSIA",
  year =         "2002",
  number =       "IDSIA-12-02",
  month =        "31 " # jul,
  keywords =     "genetic algorithms, genetic programming, OOPS,
                 bias-optimality, incremental optimal universal search,
                 metasearching, metalearning, self-improvement",
  URL =          "ftp://ftp.idsia.ch/pub/juergen/oops.ps.gz",
  abstract =     "We extend principles of non-incremental universal
                 search to build a novel, optimally fast, incremental
                 learner that is able to improve itself through
                 experience. The Optimal Ordered Problem Solver (OOPS)
                 searches for a universal algorithm that solves each
                 task in a sequence of tasks. It continually organises
                 and exploits previously found solutions to earlier
                 tasks, efficiently searching not only the space of
                 domain-specific algorithms, but also the space of
                 search algorithms.

                 The initial bias is embodied by a task-dependent
                 probability distribution on possible program prefixes
                 (pieces of code that may continue). Prefixes are
                 self-delimiting and executed in online fashion while
                 being generated. They compute the probabilities of
                 their own possible continuations. Let p^n denote a
                 found prefix solving the first n tasks. It may exploit
                 previous solutions p^i (i

Genetic Programming entries for Jurgen Schmidhuber

Citations