Cooperative Versus Competitive Coevolution for Pareto Multiobjective Optimization

AbstrakKini, semakin ramai penyelidik telah menunjukkan minat mengkaji permainan Kecerdasan Buatan (KB).Permainan seumpama ini menyediakan tapak uji yang sangat berguna dan baik untuk mengkaji asasdan teknik-teknik KB. Teknik KB, seperti pembelajaran, pencarian dan perencanaan digunakan untukmenghasilkan agen maya yang mampu berfikir dan bertindak sewajarnya dalam persekitaran permainanyang kompleks dan dinamik. Dalam kajian ini, satu set pengawal permainan autonomi untuk pasukan hantudalam permainan Ms. Pac-man yang dicipta dengan menggunakan penghibridan Evolusi PengoptimumanMultiobjektif (EPM) dan ko-evolusi persaingan untuk menyelesaikan masalah pengoptimuman dua objektifiaitu meminimumkan mata dalam permainan dan bilangan neuron tersembunyi di dalam rangkaianneural buatan secara serentak. Arkib Pareto Evolusi Strategi (APES) digunakan, teknik pengoptimumanmultiobjektif ini telah dibuktikan secara saintifik antara yang efektif di dalam pelbagai aplikasi. Secarakeseluruhannya, keputusan eksperimen menunjukkan bahawa teknik pengoptimuman multiobjektif bolehmendapat manfaat daripada aplikasi ko-evolusi persaingan Abstract Recently, researchers have shown an increased interest in game Artificial Intelligence (AI). Gamesprovide a very useful and excellent testbed for fundamental AI research. The AI techniques, such aslearning, searching and planning are applied to generate the virtual creatures that are able to think andact appropriately in the complex and dynamic game environments. In this study, a set of autonomousgame controllers for the ghost team in the Ms. Pac-man game are created by using the hybridizationof Evolutionary Multiobjective Optimization (EMO) and competitive coevolution to solve the bi-objectiveoptimization problem of minimizing the game's score by eating Ms. Pac-man agent and the number ofhidden neurons in neural network simultaneously. The Pareto Archived Evolution Strategy (PAES) is usedthat has been proved to be an effective and efficient multiobjective optimization technique in variousapplications. Overall, the results show that multiobjective optimizer can benefit from the application ofcompetitive coevolutionary

Download Full-text

Classification as Clustering: A Pareto Cooperative-Competitive GP Approach

Evolutionary Computation ◽

10.1162/evco_a_00016 ◽

2011 ◽

Vol 19 (1) ◽

pp. 137-166 ◽

Cited By ~ 8

Author(s):

Andrew R. McIntyre ◽

Malcolm I. Heywood

Keyword(s):

Multiobjective Optimization ◽

Genetic Programming ◽

Natural Environment ◽

Team Member ◽

Population Based ◽

Classification Performance ◽

Model Complexity ◽

Good Balance ◽

Competitive Coevolution ◽

Do So

Intuitively population based algorithms such as genetic programming provide a natural environment for supporting solutions that learn to decompose the overall task between multiple individuals, or a team. This work presents a framework for evolving teams without recourse to prespecifying the number of cooperating individuals. To do so, each individual evolves a mapping to a distribution of outcomes that, following clustering, establishes the parameterization of a (Gaussian) local membership function. This gives individuals the opportunity to represent subsets of tasks, where the overall task is that of classification under the supervised learning domain. Thus, rather than each team member representing an entire class, individuals are free to identify unique subsets of the overall classification task. The framework is supported by techniques from evolutionary multiobjective optimization (EMO) and Pareto competitive coevolution. EMO establishes the basis for encouraging individuals to provide accurate yet nonoverlaping behaviors; whereas competitive coevolution provides the mechanism for scaling to potentially large unbalanced datasets. Benchmarking is performed against recent examples of nonlinear SVM classifiers over 12 UCI datasets with between 150 and 200,000 training instances. Solutions from the proposed coevolutionary multiobjective GP framework appear to provide a good balance between classification performance and model complexity, especially as the dataset instance count increases.

Download Full-text