Genetic Programming Theory and Practice VI (Genetic and Evolutionary Computation) (v. 6)
Format: PDF / Kindle (mobi) / ePub
Genetic Programming Theory and Practice VI was developed from the sixth workshop at the University of Michigan’s Center for the Study of Complex Systems to facilitate the exchange of ideas and information related to the rapidly advancing field of Genetic Programming (GP). Contributions from the foremost international researchers and practitioners in the GP arena examine the similarities and differences between theoretical and empirical results on real-world problems. The text explores the synergy between theory and practice, producing a comprehensive view of the state of the art in GP application.
These contributions address several significant interdependent themes which emerged from this year’s workshop, including: (1) Making efficient and effective use of test data. (2) Sustaining the long-term evolvability of our GP systems. (3) Exploiting discovered subsolutions for reuse. (4) Increasing the role of a Domain Expert.
industry to develop molecular diagnostics (firstname.lastname@example.org). Tina Yu is an Associate Professor of Computer Science at the Memorial University of Newfoundland, Canada (email@example.com). Ren-Jie Zeng is an Assistant Research Fellow at the AI-ECON Research Center, National Chengchi University, Taiwan (firstname.lastname@example.org). Preface The work described in this book was first presented at the Sixth Workshop on Genetic Programming, Theory and Practice, organized by the Center for the Study of
practitioner is frequently left with an implicit experimental design problem. The non-deterministic methodology of GP requires that evaluation be conducted over multiple runs per data partition. Conversely, deterministic machine learning algorithms only require one evaluation per partition (for a given selection of learning parameters). The evaluation design problem of interest here therefore comes down to establishing how the single performance point from the deterministic algorithm may be
model results in an increasing computational overhead as the training exemplar count increases, where this is also reflected in the strong correlation between SVM model, complexity and size of the training data partition. 5. Conclusions Advances in areas such as Evolutionary Multi-objective Optimization (EMO) and competitive coevolution are providing new opportunities for addressing credit assignment in GP. Specifically, the proposed CMGE model scales to large data sets using a competitive
theories is correct) or whether there are multiple causes (in which case, perhaps, different theories capture different aspects of the bloat phenomenon). Elitism Elitism is a commonly used technique where one or more of the highestfitness individuals are copied, unchanged, from one generation to the next. The amount of elitism used is often characterised by the elite fraction, which is the ratio N/M between the elite size, N , and the population size, M . Elitism is typically used in
iteration. The contour plot at the right corresponds to the original response surface (only two out of 10 input variables are significant). For this adaptive DOE run we started with a base set of 12 sample points randomly distributed in a 10-D space. For each iteration, three new samples were collected: (1) at the point of maximum ensemble disagreement, (2) the model maximum and (3) the model minimum. Using this additional targeted data collection, the adaptive DOE process was able to quickly