On the editing rate of the MULTIDIT algorithm
Pattern Recognition Letters
Planning for conjunctive goals
Artificial Intelligence
A parallel network that learns to play backgammon
Artificial Intelligence
Classifier systems and genetic algorithms
Artificial Intelligence
Using local models to control movement
Advances in neural information processing systems 2
Adaptation in natural and artificial systems
Adaptation in natural and artificial systems
Practical Issues in Temporal Difference Learning
Machine Learning
A teaching method for reinforcement learning
ML92 Proceedings of the ninth international workshop on Machine learning
Selecting typical instances in instance-based learning
ML92 Proceedings of the ninth international workshop on Machine learning
Tolerating noisy, irrelevant and novel attributes in instance-based learning algorithms
International Journal of Man-Machine Studies - Special issue: symbolic problem solving in noisy and novel task environments
Reinforcement learning for the adaptive control of perception and action
Reinforcement learning for the adaptive control of perception and action
Training agents to perform sequential behavior
Adaptive Behavior
Robot shaping: developing autonomous agents through learning
Artificial Intelligence
Genetic Algorithms in Search, Optimization and Machine Learning
Genetic Algorithms in Search, Optimization and Machine Learning
Learning Sequential Decision Rules Using Simulation Models and Competition
Machine Learning - Special issue on genetic algorithms
Learning to Predict by the Methods of Temporal Differences
Machine Learning
Distance Metrics for Instance-Bsed Learning
ISMIS '91 Proceedings of the 6th International Symposium on Methodologies for Intelligent Systems
Colearning in Differential Games
Machine Learning
Accelerating reinforcement learning by composing solutions of automatically identified subtasks
Journal of Artificial Intelligence Research
Hi-index | 0.00 |
Combining different machine learning algorithms in the same system canproduce benefits above and beyond what either method could achievealone. This paper demonstrates that genetic algorithms can be used inconjunction with lazy learning to solve examples of a difficult class ofdelayed reinforcement learning problems better than either method alone.This class, the class of differential games,includes numerous important control problems that arise in robotics,planning, game playing, and other areas, and solutions for differentialgames suggest solution strategies for the general class of planning and control problems. We conducted a seriesof experiments applying three learning approaches – lazy Q-learning,k-nearest neighbor (k-NN), and a genetic algorithm – to aparticular differential game called a pursuit game. Our experimentsdemonstrate that k-NN had great difficulty solving the problem, while alazy version of Q-learning performed moderately well andthe genetic algorithm performed even better. These resultsmotivated the next step in the experiments, where we hypothesizedk-NN was having difficulty because it did not have good examples – acommon source of difficulty for lazy learning. Therefore, we used thegenetic algorithm as a bootstrapping method for k-NN to createa system to provide these examples. Our experimentsdemonstrate that the resulting joint system learned to solve thepursuit games with a high degree of accuracy – outperforming eithermethod alone – and with relatively small memory requirements.