Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/springer-journals/optimal-ordered-problem-solver-z06QCs7hvq
References (131)
-
A. Newell, H. Simon (1963)
Computers and Thought -
R. Solomonoff (1964)
A formal theory of inductive inferenceInformation and Control, 7
-
the n -th frozen program (either user-defined or frozen by oops ) stored in q below a frozen , assuming (somewhat arbitrarily) zero inputs and outputs -
R. Solomonoff (1989)
A SYSTEM FOR INCREMENTAL LEARNING BASED ON ALGORITHMIC PROBABILITY -
P. Utgoff (1984)
Shift of bias for inductive concept learning -
A stack of integer arrays, each having a name, an address, and a size (not used in this paper, but implemented and mentioned for the sake of completeness) -
E. B. Baum, I. Durdanovic (1999)
Toward a model of mind as an economy of agentsMachine Learning, 35
-
Jürgen Schmidhuber (2002)
The Speed Prior: A New Simplicity Measure Yielding Near-Optimal Computable Predictions -
M. Tsukamoto (1977)
Program Stacking Technique, 17
-
A variable holding the index r of this tape’s task -
R. Waldinger, Richard Lee (1969)
PROW: A Step Toward Automatic Program Writing -
At time 3 ∗ 10 9 -
which defines self-made recursive code functionally equivalent to fac(n), which calls itself by calling the most recent self-made function even before it is completely defined -
For the 1 n 2 n -problem, within 480,000 time steps (less than a second), oops found nongeneral but working code for n = 1: (defnp 2toD) -
A. Newell, H. Simon (1995)
GPS, a program that simulates human thought -
by time 541 ∗ 10 9 (roughly 3 days), it had found fresh code (new a last ) for n = 1 , 2 , 3: (c3 dec boostq defnp c4 calltp c3 c5 calltp endnp) -
J. Schmidhuber (1993)
Proceedings of the International Conference on Artificial Neural Networks -
J. Schmidhuber, Jieyu Zhao, M. Wiering (1996)
Simple Principles of Metalearning -
A. Turing (1937)
On computable numbers, with an application to the EntscheidungsproblemProc. London Math. Soc., s2-42
-
J. Schmidhuber (1993)
Proc. of the Intl. Conf. on Artificial Neural Networks -
Marcus Hutter (2002)
Self-Optimizing and Pareto-Optimal Policies in General Environments based on Bayes-MixturesArXiv, cs.AI/0204040
-
P. Werbos (1974)
Beyond Regression : "New Tools for Prediction and Analysis in the Behavioral Sciences -
M. Dorigo, G. Caro, L. Gambardella (1999)
Ant Algorithms for Discrete OptimizationArtificial Life, 5
-
A binary quoteflag determining whether the instructions pointed to by ip will get executed or just quoted , that is, pushed onto ds -
C. Anderson (1986)
Learning and problem-solving with multilayer connectionist systems (adaptive, strategy learning, neural networks, reinforcement learning) -
An auxiliary data stack Ds -
David Bulman (1977)
Stack Computers [Guest editor's introduction]Computer, 10
-
H. P. Schwefel (1974)
Numerische optimierung von computer-modellen -
(1993)
The SOAR Papers -
(2002)
OPTIMAL ORDERED PROBLEM SOLVER -
D. Wolpert, W. Macready (1995)
No Free Lunch Theorems for Search -
J. Schmidhuber (2003)
Advances in Neural Information Processing Systems 15 -
J. Schmidhuber (2000)
Algorithmic Theories of EverythingArXiv, quant-ph/0011122
-
G. Chaitin (1975)
A Theory of Program Size Formally Identical to Information TheoryJ. ACM, 22
-
J. Schmidhuber (2003)
Real AI: New Approaches to Artificial General Intelligence -
J. Schmidhuber, Viktor Zhumatiy, M. Gagliolo, F. Groen, N. Amato, Andrea Bonarini, E. Yoshida, B. Kröse (2004)
Bias-Optimal Incremental Learning of Control Sequences for Virtual Robots -
E. Baum (1999)
Toward a Model of Intelligence as an Economy of AgentsMachine Learning, 35
-
I. Rechenberg (1973)
Evolutionsstrategie : Optimierung technischer Systeme nach Prinzipien der biologischen Evolution -
J. Schmidhuber (2002)
Hierarchies of Generalized Kolmogorov Complexities and Nonenumerable Universal Measures Computable in the LimitInt. J. Found. Comput. Sci., 13
-
K. Zuse (1991)
Rechnender Raum -
D. Rumelhart, Geoffrey Hinton, Ronald Williams (1986)
Learning internal representations by error propagation -
S. Hochreiter, A. Younger, P. Conwell (2001)
Learning to Learn Using Gradient Descent -
S. Hochreiter, A. S. Younger, P. R. Conwell (2001)
Proc. Intl. Conf. on Artificial Neural Networks (ICANN-2001) -
P. Utgoff (1986)
Machine Learning -
J. Schmidhuber (1993)
An 'introspective' network that can learn to run its own weight change algorithm -
J. SCHMIDHUBER -
J. Schmidhuber (1997)
Discovering Neural Nets with Low Kolmogorov Complexity and High Generalization CapabilityNeural networks : the official journal of the International Neural Network Society, 10 5
-
R. Solomonoff (1985)
The Application of Algorithmic Probability to Problems in Artificial Intelligence -
inc(x) returns x + 1; dec(x) returns x − 1; by2(x) returns 2x; add(x,y) returns x + -
Simple Stack Manipulators. del() decrements dp; clear() sets dp := 0; dp2ds() returns dp; setdp(x) sets dp := x; ip2ds( -
J. Holland (1985)
Properties of the Bucket Brigade -
Ming Li, P. Vitányi (1993)
An Introduction to Kolmogorov Complexity and Its Applications -
H. Bremermann (1982)
Minimum energy requirements of information transfer and computingInternational Journal of Theoretical Physics, 21
-
R. Salustowicz, J. Schmidhuber (1998)
Evolving Structured Programs with Hierarchical Instructions and Skip Nodes -
J. Schmidhuber, Jieyu Zhao, M. Wiering (1997)
Shifting Inductive Bias with Success-Story Algorithm, Adaptive Levin Search, and Incremental Self-ImprovementMachine Learning, 28
-
Instruction jmp1(val, n) sets cs [ cp ] -
Wolfgang Banzhaf, Peter Nordin, Robert Keller, F. Francone (1997)
Genetic Programming: An Introduction -
(2003)
Towards solving the grand problem of AI -
J. H. Holland (1985)
Proceedings of an International Conference on Genetic Algorithms -
(1974)
Numerische Optimierung von Computer-Modellen. Dissertation -
J. Schmidhuber (1993)
A ‘Self-Referential’ Weight Matrix -
A data stack ds(r ) (or ds for short, omitting the task index) for storing function arguments. (The corresponding stack pointer is dp : 0 ≤ dp ≤ maxdp) -
(1966)
Advances in Neural Information Processing Systems 4 (pp. 831–838) -
(2001)
General methods for search and reinforcement learning. Application for SNF grant 21-6678 -
C. Green (1969)
Application of Theorem Proving to Problem Solving -
Operand and(x,y) returns 1 if x > 0 and y > 0, otherwise 0. Analogously for or(x,y) -
1, 1, fac , up c1 ex rt0 del up dec fac mul ret) declares a recursive function fac (n) which returns 1 if n = 0, otherwise returns n × fac (n-1) -
J. Schmidhuber, Jieyu Zhao, N. Schraudolph (1998)
Reinforcement Learning with Self-Modifying Policies -
J. Schmidhuber (2002)
Proceedings of the 15th Annual Conference on Computational Learning Theory (COLT 2002) -
J. Koopman (1989)
Stack computers: the new wave -
Jürgen Schmidhuber (1987)
Evolutionary principles in self-referential learning, or on learning how to learn: The meta-meta-. hook -
N. L. Cramer (1985)
Proceedings of an International Conference on Genetic Algorithms and Their Applications -
(1994)
Artificial Intelligence: A Modern Approach -
L. Gambardella, M. Dorigo (2000)
An Ant Colony System Hybridized with a New Local Search for the Sequential Ordering ProblemINFORMS J. Comput., 12
-
works like cpn(1); cpnb(n) copies n ds entries above ds -
Jana Koehler, Bernhard Nebel, J. Hoffmann, Yannis Dimopoulos (1997)
Extending Planning Graphs to an ADL Subset -
Charles Bennett (1982)
The thermodynamics of computation—a reviewInternational Journal of Theoretical Physics, 21
-
A. Kolmogorov (1965)
Three approaches to the quantitative definition of informationProblems of Information Transmission, 1
-
P. Langley (1985)
Learning to Search: From Weak Methods to Domain-Specific HeuristicsCogn. Sci., 9
-
S. Kothari, H. Oh (1993)
Neural Networks for Pattern RecognitionAdv. Comput., 37
-
A.1 Data Structures on Tapes Each tape r contains various stack-like data structures represented as sequences of integers -
At time 10 7 (roughly 10 s ) it had solved the 2nd instance by simply prolonging the previous code, using the old, unchanged start address a last : (defnp 2toD grt c2 c2 endnp) -
(1973)
Universal sequential search problems -
At time 2 . 85 ∗ 10 9 (less than 1 hour) it had solved the 4th instance through prolongation: (defnp 2toD grt c2 c2 endnp boostq delD delD bsf 2toD fromD delD delD delD fromD bsf by2 bsf) -
sub(x,y) returns x − y; mul(x,y) returns x * y; div(x,y) returns the smallest integer ≤ x/y; powr(x,y) returns x y (and costs y unit time steps) -
J. Schmidhuber (1995)
Machine Learning: Proceedings of the Twelfth International Conference -
Then oops switched to the Hanoi problem. Almost immediately (less than 1 ms later), at time 30 , 665 , 064 , 995, it had found the trivial code for n = 1: ( -
A stack pats of search patterns -
R. Salustowicz, M. Wiering, J. Schmidhuber (1998)
Learning Team Strategies: Soccer Case StudiesMachine Learning, 33
-
S. Wolfram (1983)
Universality and complexity in cellular automataPhysica D: Nonlinear Phenomena, 10
-
J. Schmidhuber (2002)
Bias-Optimal Incremental Problem Solving -
L. Levin (1984)
Randomness Conservation Inequalities; Information and Independence in Mathematical TheoriesInf. Control., 61
-
R. P. Salustowicz, J. Schmidhuber (1998)
Machine Learning: Proceedings of the Fifeteenth International Conference (ICML'98) -
Michael Jordan, D. Rumelhart (1992)
Forward Models: Supervised Learning with a Distal TeacherCogn. Sci., 16
-
J. Schmidhuber (2001)
Sequential Decision Making Based on Direct Search -
Marcus Hutter (2000)
Towards a Universal Theory of Artificial Intelligence Based on Algorithmic Probability and Sequential DecisionsArXiv, cs.AI/0012011
-
(1987)
Learning how the world works: Specifications for predictive networks in robots and brains -
A. Kolmogorov (1968)
Three approaches to the quantitative definition of informationInternational Journal of Computer Mathematics, 2
-
(1950)
Random processes and transformations -
L. Kaelbling, M. Littman, A. Moore (1996)
Reinforcement learning: A surveyJournal of AI research, 4
-
Ivo Kwee, Marcus Hutter, J. Schmidhuber (2001)
Market-Based Reinforcement Learning in Partially Observable Worlds -
後藤 滋樹 (2002)
20世紀の名著名論：A.M.Turing: On Computable Numbers with an Application to the Entscheidungsproblem, 43
-
J. Neumann, A. Burks (1967)
Theory Of Self Reproducing Automata -
S. Lloyd (1999)
Ultimate physical limits to computationNature, 406
-
N. Cramer (1985)
A Representation for the Adaptive Generation of Simple Sequential Programs -
Y. Deville, K. Lau (1994)
Logic Program SynthesisJ. Log. Program., 19/20
-
W. Vent (1975)
Rechenberg, Ingo, Evolutionsstrategie — Optimierung technischer Systeme nach Prinzipien der biologischen Evolution. 170 S. mit 36 Abb. Frommann‐Holzboog‐Verlag. Stuttgart 1973. BroschiertFeddes Repertorium, 86
-
(1987)
Der genetische Algorithmus: Eine Implementierung in Prolog. Fortgeschrittenenpraktikum, Institut für Informatik, Lehrstuhl Prof. Radig, Technische Universität München -
L. Kaelbling, M. Littman, A. Moore (1996)
Reinforcement Learning: A SurveyJ. Artif. Intell. Res., 4
-
J. Schmidhuber (2003)
Soft Computing and complex systems -
R. Olsson (1995)
Inductive Functional Programming Using Incremental Program TransformationArtif. Intell., 74
-
J. Schmidhuber (2003)
Exploring the predictable -
R. Solomonoff (1964)
A Formal Theory of Inductive Inference. Part IIInf. Control., 7
-
D. Lenat (1983)
Theory Formation by Heuristic SearchArtif. Intell., 21
-
W. Banzhaf, F. Francone, Robert Keller, P. Nordin (1998)
Genetic programming - An Introduction: On the Automatic Evolution of Computer Programs and Its Applications -
E. Fredkin, T. Toffoli (2002)
Conservative logicInternational Journal of Theoretical Physics, 21
-
J. Schmidhuber (2003)
Goedel Machines: Self-Referential Universal Problem Solvers Making Provably Optimal Self-ImprovementsArXiv, cs.LO/0309048
-
J. Schmidhuber (1994)
On learning how to learn learning strategies -
V. Vapnik (1991)
Principles of Risk Minimization for Learning Theory -
J. Schmidhuber (1990)
Reinforcement Learning in Markovian and Non-Markovian Environments -
W. Gasarch (1997)
Book Review: An introduction to Kolmogorov Complexity and its Applications Second Edition, 1997 by Ming Li and Paul Vitanyi (Springer (Graduate Text Series))ACM SIGACT News, 28
-
K. Gödel (1931)
Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme IMonatshefte für Mathematik, 149
-
by time 260 ∗ 10 9 (more than 1 day), it had found fresh yet somewhat bizarre code (new start address a last ) for n = 1 , 2: (c4 c3 cpn c4 by2 c3 by2 exec) -
J. Holland (1975)
Adaptation in natural and artificial systems -
Marcus Hutter (2000)
The Fastest and Shortest Algorithm for all Well-Defined ProblemsArXiv, cs.CC/0206022
-
M. Wiering, J. Schmidhuber (1996)
Solving POMDPs with Levin Search and EIRA -
C. Moore, G. Leach (1970)
Forth - a language for interactive computing -
D. Nguyen, B. Widrow (1989)
The truck backer-upper: an example of self-learning in neural networksInternational 1989 Joint Conference on Neural Networks
-
(1974)
Laws of information (nongrowth) and aspects of the foundation of probability theory -
R. Salustowicz, Jürgen Schmidhuber (1997)
Probabilistic Incremental Program EvolutionEvolutionary Computation, 5
-
Corso Elvezia (1995)
Discovering Solutions with Low Kolmogorov Complexity and High Generalization Capability
- Publisher
- Springer Journals
- Copyright
- Copyright © 2004 by Kluwer Academic Publishers
- Subject
- Computer Science; Artificial Intelligence (incl. Robotics); Control, Robotics, Mechatronics; Computing Methodologies; Simulation and Modeling; Language Translation and Linguistics
- ISSN
- 0885-6125
- eISSN
- 1573-0565
- DOI
- 10.1023/B:MACH.0000015880.99707.b2
- Publisher site
- See Article on Publisher Site
Abstract
We introduce a general and in a certain sense time-optimal way of solving one problem after another, efficiently searching the space of programs that compute solution candidates, including those programs that organize and manage and adapt and reuse earlier acquired knowledge. The Optimal Ordered Problem Solver (OOPS) draws inspiration from Levin's Universal Search designed for single problems and universal Turing machines. It spends part of the total search time for a new problem on testing programs that exploit previous solution-computing programs in computable ways. If the new problem can be solved faster by copy-editing/invoking previous code than by solving the new problem from scratch, then OOPS will find this out. If not, then at least the previous solutions will not cause much harm. We introduce an efficient, recursive, backtracking-based way of implementing OOPS on realistic computers with limited storage. Experiments illustrate how OOPS can greatly profit from metalearning or metasearching, that is, searching for faster search procedures.
Journal
Machine Learning – Springer Journals
Published: Oct 18, 2004