Optimal Experiment Design for Coevolutionary Active Learning

This paper presents a policy for selecting the most informative individuals in a teacher-learner type coevolution. We propose the use of the surprisal of the mean, based on Shannon information theory, which best disambiguates a collection of arbitrary and competing models based solely on their predictions. This policy is demonstrated within an iterative coevolutionary framework consisting of symbolic regression for model inference and a genetic algorithm for optimal experiment design. Complex symbolic expressions are reliably inferred using fewer than 32 observations. The policy requires 21% fewer experiments for model inference compared to the baselines and is particularly effective in the presence of noise corruption, local information content as well as high dimensional systems. Furthermore, the policy was applied in a real-world setting to model concrete compression strength, where it was able to achieve 96.1% of the passive machine learning baseline performance with only 16.6% of the data.

[1]  Hod Lipson,et al.  Automated reverse engineering of nonlinear dynamical systems , 2007, Proceedings of the National Academy of Sciences.

[2]  Hod Lipson,et al.  Active Coevolutionary Learning of Deterministic Finite Automata , 2005, J. Mach. Learn. Res..

[3]  Ran El-Yaniv,et al.  Online Choice of Active Learning Algorithms , 2003, J. Mach. Learn. Res..

[4]  Sean Luke,et al.  Cooperative Multi-Agent Learning: The State of the Art , 2005, Autonomous Agents and Multi-Agent Systems.

[5]  Michel Gevers.  Identification for Control: From the Early Achievements to the Revival of Experiment Design , 2005, CDC 2005.

[6]  Slavko Simic.  Best possible global bounds for Jensen's inequality , 2009, Appl. Math. Comput..

[7]  Dick den Hertog,et al.  Order of Nonlinearity as a Complexity Measure for Models Generated by Symbolic Regression via Pareto Genetic Programming , 2009, IEEE Transactions on Evolutionary Computation.

[8]  Hod Lipson,et al.  Predicting solution rank to improve performance , 2010, GECCO '10.

[9]  David A. Cohn,et al.  Improving generalization with active learning , 1994, Machine Learning.

[10]  Araceli Sanchis,et al.  A general learning co-evolution method to generalize autonomous robot navigation behavior , 2000, Proceedings of the 2000 Congress on Evolutionary Computation. CEC00 (Cat. No.00TH8512).

[11]  Graham C. Goodwin,et al.  Robust optimal experiment design for system identification , 2007, Autom..

[12]  Malcolm I. Heywood,et al.  GP Classification under Imbalanced Data sets: Active Sub-sampling and AUC Approximation , 2008, EuroGP.

[13]  Rasoul Karimi,et al.  Active Learning for Recommender Systems , 2015, KI - Künstliche Intelligenz.

[14]  Hod Lipson,et al.  Distilling Free-Form Natural Laws from Experimental Data , 2009, Science.

[15]  Hod Lipson,et al.  Coevolution of Fitness Predictors , 2008, IEEE Transactions on Evolutionary Computation.

[16]  Aleksey V. Nenarokomov,et al.  Uncertainties in parameter estimation: the optimal experiment design , 2000 .

[17]  Hod Lipson,et al.  Nonlinear system identification using coevolution of models and tests , 2005, IEEE Transactions on Evolutionary Computation.

[18]  Hod Lipson,et al.  Resilient Machines Through Continuous Self-Modeling , 2006, Science.

[19]  Stéphane Doncieux,et al.  Automatic system identification based on coevolution of models and tests , 2009, 2009 IEEE Congress on Evolutionary Computation.

[20]  Hod Lipson,et al.  Trainer selection strategies for coevolving rank predictors , 2011, 2011 IEEE Congress of Evolutionary Computation (CEC).

[21]  Hod Lipson,et al.  Automating Genetic Network Inference with Minimal Physical Experimentation Using Coevolution , 2004, GECCO.

[22]  Samir W. Mahfoud.  Niching methods for genetic algorithms , 1996 .

[23]  I-Cheng Yeh,et al.  Analysis of Strength of Concrete Using Design of Experiments and Neural Networks , 2006 .

[24]  Lennart Ljung,et al.  Some results on optimal experiment design , 2000, Autom..

[25]  Andreas Krause,et al.  Near-optimal sensor placements in Gaussian processes , 2005, ICML.

[26]  H. Wynn,et al.  Maximum entropy sampling and optimal Bayesian experimental design , 2000 .

[27]  I-Cheng Yeh,et al.  Modeling of strength of high-performance concrete using artificial neural networks , 1998 .

[28]  Xavier Bombois,et al.  Least costly identification experiment for control , 2006, Autom..

[29]  Burr Settles,et al.  Active Learning Literature Survey , 2009 .

[30]  John R. Koza,et al.  Genetic programming - on the programming of computers by means of natural selection , 1993, Complex adaptive systems.

[31]  Carl E. Rasmussen,et al.  Learning to Control a Low-Cost Manipulator using Data-Efficient Reinforcement Learning , 2011, Robotics: Science and Systems.

[32]  David A. Cohn,et al.  Active Learning with Statistical Models , 1996, NIPS.

[33]  Edwin D. de Jong,et al.  Ideal Evaluation from Coevolution , 2004, Evolutionary Computation.