Learning-and-Evolution.md

Learning and Evolution

For an early review regarding to this topic, please refer to Nolfi and Floreano's 1999 paper on Autonomous Robots.

Benefits and Costs

In Nolfi and Floreano's 1999 paper,

Multi-Scale/Level Evolution and Learning

"The processes of life involve change at many scales of space and time. Choosing a spatiotemporal scale emphasizes the change at that scale, rendering smaller scales essentially as noise, large scales essentially as constant. One trade-off between learning (at the individual level) and evolution (at the level of population or species) is readily apparent: Learning is facilitated by long individual lifetimes, whereas evolution benefits from rapidly passing generations." [Ackley&Littman, 1991].

Baldwin Effect

As pointed out by Hinton (ACM Turing Award Laureate) and Nowlan, "The main limitation of the Baldwin effect is that it is only effective in spaces that would be hard to search without an adaptive process to restrcuture the space".

"acquired characters"

In Whitley et al.'s paper on PPSN-1994, initial simulation experiments demonstrated that GA with the Baldwin effect sometimes finds the global optimum while Lamarckian GA sometimes gets trapped into the local optima, though the latter shows faster convergence at the early evolution stage. However, more simulation and real-life experiments are still expected nowadays to compare them. Recently, Todd (from Stanford University) etc. have investigated the effects of different design choices of learning and evolution on the overall performance on NK fitness landscapes.

Reference

• Levin, M., 2023. Darwin’s agential materials: Evolutionary implications of multiscale competency in developmental biology. Cellular and Molecular Life Sciences, 80(6), p.142.
• Miconi, T., 2023, July. Learning to acquire novel cognitive tasks with evolution, plasticity and meta-meta-learning. In International Conference on Machine Learning (pp. 24756-24774). PMLR.
• Pigozzi, F., Camerota Verdù, F.J. and Medvet, E., 2023, July. How the morphology encoding influences the learning ability in body-brain co-optimization. In Proceedings of Genetic and Evolutionary Computation Conference (pp. 1045-1054). ACM.
• Gunawardena, J., 2022. Learning outside the brain: Integrating cognitive science and systems biology. Proceedings of the IEEE, 110(5), pp.590-612.
• Todd, G., Candadai, M. and Izquierdo, E.J., 2020, July. Interaction between evolution and learning in NK fitness landscapes. In Proceedings of Artificial Life Conference (pp. 761-767). MIT Press.
• Bartoli, A., Catto, M., De Lorenzo, A., Medvet, E. and Talamini, J., 2020, July. Mechanisms of social learning in evolved artificial life. In Proceedings of Artificial Life Conference (pp. 190-198). MIT Press.
• Fernando, C., Banarse, D., Blundell, C., Zwols, Y., Ha, D., Rusu, A.A., Pritzel, A. and Wierstra, D., 2017. Pathnet: Evolution channels gradient descent in super neural networks. arXiv preprint arXiv:1701.08734.
• Barkoczi, D. and Galesic, M., 2016. Social learning strategies modify the effect of network structure on group performance. Nature Communications, 7(1), p.13109.
• Nolfi, S. and Floreano, D., 1999. Learning and evolution. Autonomous Robots, 7, pp.89-113.
• Whitley, D., Gordon, V.S. and Mathias, K., 1994. Lamarckian evolution, the Baldwin effect and function optimization. In International Parallel Problem Solving from Nature (pp. 5-15). Springer.
• Nolfi, S., Parisi, D. and Elman, J.L., 1994. Learning and evolution in neural networks. Adaptive Behavior, 3(1), pp.5-28.
• Gruau, F. and Whitley, D., 1993. Adding learning to the cellular development of neural networks: Evolution and the Baldwin effect. Evolutionary Computation, 1(3), pp.213-233.
• Vose, M.D. and Liepins, G.E., 1991. Punctuated equilibria in genetic search. Complex Systems, 5(1), pp.31-44.
• Ackley, D. and Littman, M., 1991. Interactions between learning and evolution. Artificial Life, 10, pp.487-509.
• Belew, R.K., 1990. Evolution, learning and culture: computational metaphors for adaptive search. Complex Systems, 4, pp.11-49.
• Belew, R.K., 1989, June. When both individuals and populations search: Adding simple learning to the genetic algorithm. In Proceedings of International Conference on Genetic Algorithms (pp. 34-41).
• Kauffman, S. and Levin, S., 1987. Towards a general theory of adaptive walks on rugged landscapes. Journal of Theoretical Biology, 128(1), pp.11-45.
• Hinton, G.E. and Nowlan, S.J., 1987. How learning can guide evolution. Complex Systems, 1(3), pp.495-502.
• Waddington, C.H., 1942. Canalization of development and the inheritance of acquired characters. Nature, 150(3811), pp.563-565.
• Baldwin, J.M., 1896. A new factor in evolution. American Naturalist, 30(355), pp.536-553.