Pareto analysis evolutionary and learning systems

Abstract

This paper attempts to argue that most adaptive systems, such as evolutionary or learning systems, have inherently multiple objectives to deal with. Very often, there is no single solution that can optimize all the objectives. In this case, the concept of Pareto optimality is key to analyzing these systems.

To support this argument, we first present an example that considers the robustness and evolvability trade-off in a redundant genetic representation for simulated evolution. It is well known that robustness is critical for biological evolution, since without a sufficient degree of mutational robustness, it is impossible for evolution to create new functionalities. On the other hand, the genetic representation should also provide the chance to find new phenotypes, i.e., the ability to innovate. This example shows quantitatively that a trade-off between robustness and innovation does exist in the studied redundant representation.

Interesting results will also be given to show that new insights into learning problems can be gained when the concept of Pareto optimality is applied to machine learning. In the first example, a Pareto-based multi-objective approach is employed to alleviate catastrophic forgetting in neural network learning. We show that learning new information and memorizing learned knowledge are two conflicting objectives, and a major part of both information can be memorized when the multi-objective learning approach is adopted. In the second example, we demonstrate that a Pareto-based approach can address neural network regularizationmore elegantly. By analyzing the Pareto-optimal solutions, it is possible to identifying interesting solutions on the Pareto front.

Keywords

Pareto analysis / multi-objective optimization / evolution / evolvability / robustness / learning / accuracy / complexity

References

[1] Ancel L, Bull J. Fighting change with change: adaptive variation in an uncertain world. Trends in Ecology and Evolution, 2002, 17(12): 551-557

[2] Tononi G, Sporns O, Edelman G. A measure for brain complexity: relating functional segregation and integration in the nervous system. In: Proceedings of the National Academy of Science of the United States of America, 1994, 91: 5033-5031

[3] Teo J, Abbass H. Multiobjectivity and complexity in embodied cognition. IEEE Transactions on Evolutionary Computation, 2005, 9(4): 337-360

[4] Louis S, Rawlines G. Pareto optimality, GA-easiness and deception. In: Proceedings of The Fifth International Conference on Genetic Algorithms. Morgan Kaufmann, 1993, 118-123

[5] Knowles J, Watson R, Corne D. Reducing local optima in singleobjective problems by multi-objectivization. In: Proceedings of International Conference on Evolutionary Muti-Griterion Optimization. Berlin: Springer, LNCS, 2001, 1993: 269-283

[6] Jensen M. Helper-objectives: using multi-objective evolutionary algorithms for single-objective optimization. Journal of Mathematical Modeling and Algorithms, 2004, 3(4): 323-347

[7] Bui L, Branke J, Abbass H. Multi-objective optimization for dynamic environments. In: Proceedings of Congress on Evolutionary Computation. IEEE, 2005, 2349-2356

[8] Jin Y, Sendhoff B. Pareto-based multi-objective machine learning: an overview and case studies. IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, 2008, 38(3): 397-415

[9] Allman J. Evolving Brains. Scientific American Library, 1999

[10] Jones B, Jin Y, Sendhoff B, Yao X. Evolving functional symmetry in a three dimensional model of an elongated organism. Artificial Life XI, 2008, 305-312

[11] Stearns S. Trade-offs in life-history evolution. Functional Ecology, 1989, 3: 259-268

[12] Mukhopadhyay A, Tissenbaum H. Reproduction and longevity: secrets revealed by c. elegans. Trends in Cell Biology, 2007, 17(2): 65-71

[13] Paenke I, Branke J, Jin Y. On the influence of phenotype plasticity on genotype diversity. In: Proceedings of 2007 IEEE Symposium on Foudations of Computational Intelligence (FOCI), 2007, 33-40

[14] Charnov E, Ernest S. The offspring-size / clutch-size trade-off in mammals. The American Naturalist, 2006, 167(4): 578-582

[15] Handl J, Kell D, Knowles J. Multi-objective optimization in computational biology and bioinformatics. ACM/IEEE Transactions on Computational Biology and Bioinformations, 2007, 4: 279-292

[16] Kitano H. Biological robustness. Nature Reviews Genetics, 2004, 5(11): 826-837

[17] Wagner A. Robustness and Evolvability in Living Systems. Princeton University Press, 2007

[18] Lenski R, Barrick J, Ofria C. Balancing robustness and evolvability. Public Library of Science Biology, 2006, 4(2): e428

[19] Cilibert S, Martin O, Wagner A. Innovation and robustness in complex gene networks. In: Proceedings of the National Academy of Science of the United states of America, 2007, 104(34): 13591-13596

[20] Deb K, Agrawal S, Pratab A, Meyarivan T. A fast elitist nondominated sorting genetic algorithm for multi-objective optimization: NSGA-II. In: Schoenauer M, Deb K, Rudolph G, . eds. Proceedings of the Parallel Problem Solving from Nature VI Conference. Paris: Springer, LNCS, 2000, 1917: 849-858

[21] Lehre P K, Haddow P C. Phenotypic complexity and local variations in neutral degree. Biosystems, 2007, 87: 233-242

[22] Kirschner M, Gerhart J. Evolvability. In: Proceedings of the National Academy of Science of the United states of America, 1998, 95(15): 8420-8427

[23] Fernández P, Solé R V. Neutral fitness landscapes in signalling networks. Journal of the Royal Society, Interface / the Royal Society, 2007, 4: 41-47

[24] Stadler B M R, Stadler P F, Wagner G P, Fontana W. The topology of the possible: formal spaces underlying patterns of evolutionary change. Journal of Theoretical Biology, 2000, 213: 41-47

[25] Gruna R. Analysis of redundant genotype-phenotype mappings- Investigation of the effect of neutrality on evolvability and robustness. <DissertationTip/>. Universität Karlsruhe, 2007

[26] Magurran A. Ecological Diversity and its Measurement. Princeton University Press, 1988

[27] Yu T. Program evolvability under environmental variations and neutrality. In: Proceedings of European Conference on Artificial Life (ECAL 2007), 2007, 835-844

[28] Jin Y. Fuzzy modeling of high-dimensional systems: complexity reduction and interpretability improvement. IEEE Transactions on Fuzzy Systems, 2000, 8(2): 212-221

[29] Jin Y, Advanced Fuzzy Systems Design and Applications. Heidelberg: Physica-Verlag/Springer-Verlag, 2003

[30] Igel C. Multi-objective model selection for support vector machines. In: Proceedings of Evolutionary Multi-Criterion Optimization, LNCS, 2005, 3410: 534-546

[31] Olshausen B. Relations between the statistics of narutal image and the response property of cortical cells. Nature, 1996, 381: 607-609

[32] Abraham W, Robins A. Memory retention-the synaptic stability versus plasticity dilemma. Trends in Neuroscience, 2005, 28(2): 73-78

[33] McCloskey M, Cohen N. Catastrophic interference in connectionist networks: the sequential learning problem. The Psychology of Learning and Motivation, 1989, 24: 109-165

[34] French R. Catastrophic forgetting in connectionist networks. Trends in Cognitive Sciences, 1999, 3(4): 128-135

[35] Jin Y, Sendhoff B. Alleviating catastrophic forgetting via multiobjective learning. In: Proceedings of International Joint Conference on Neural Networks, 2006, 3335-3342

RIGHTS & PERMISSIONS

2014 Higher Education Press and Springer-Verlag Berlin Heidelberg