Abstract
Learning has been shown to be beneficial to in creating more adaptive algorithms, and also in evolving neural networks. Moreover, learning can be classified into two types, namely social learning, or learning from others (e.g., imitation), and individual trial-and-error learning. A “social learning strategy” – a rule governing whether and when to use social or individual learning, is often said to be more beneficial than relying on social or individual learning alone. In this paper we compare the effect on evolution of social learning in comparison with that of individual learning. A neural architecture called a “self-taught neural network” is proposed in order to allow an agent to learn on its own, without any supervision. We simulate a multi-agent system in which agents, each controlled by a neural network, have to develop adaptive behaviour and compete with each other for survival. Experimental results show that evolved self-teaching presents the most effective behaviour in our simulated world. We conclude this paper with some indications for future work.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The term copy is often used to stand for any form of social learning, not just copying.
- 2.
The reinforcement and the action modules need not have the same topology. In our simulation, the reinforcement module possesses the same neuronal structure as the action module, but has 10 hidden neurons separate from the hidden neurons of the action module.
References
Baldwin, J.M.: A new factor in evolution. Am. Nat. 30(354), 441–451 (1896)
Le, N.: Organic selection and social heredity: the original Baldwin effect revisited. In: The 2019 Conference on Artificial Life. MIT Press (2019)
Hinton, G.E., Nowlan, S.J.: How learning can guide evolution. Complex Syst. 1, 495–502 (1987)
Ackley, D., Littman, M.: Interactions between learning and evolution. In: Langton, C.G., Taylor, C., Farmer, J.D., Rasmussen, S. (eds.) Artificial Life II. SFI Studies in the Sciences of Complexity, vol. X, pp. 487–509. Addison-Wesley, Reading (1992)
Harvey, I.: Is there another new factor in evolution? Evol. Comput. 4(3), 313–329 (1996)
Le, N.: Evolving self-taught neural networks: the Baldwin effect and the emergence of intelligence. In: 2019 AISB Annual Convention - 10th Symposium on AI & Games, Falmouth, UK, 16–18 April 2019
Nolfi, S., Parisi, D., Elman, J.L.: Learning and evolution in neural networks. Adapt. Behav. 3(1), 5–28 (1994)
Soltoggio, A., Stanley, K.O., Risi, S.: Born to learn: the inspiration, progress, and future of evolved plastic artificial neural networks. Neural Netw. 108, 48–67 (2018)
Le, N., Brabazon, A., O’Neill, M.: The evolution of self-taught neural networks in a multi-agent environment. In: Kaufmann, P., Castillo, P.A. (eds.) EvoApplications 2019. LNCS, vol. 11454, pp. 457–472. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-16692-2_31
Le, N.: Evolution and self-teaching in neural networks. In: Proceedings of the Genetic and Evolutionary Computation Conference Companion (GECCO 2019). ACM Press (2019)
Laland, K.N.: Social learning strategies. Anim. Learn. Behav. 32(1), 4–14 (2004). https://doi.org/10.3758/BF03196002
Le, N., O’Neill, M., Brabazon, A.: Adaptive advantage of learning strategies: a study through dynamic landscape. In: Auger, A., Fonseca, C.M., Lourenço, N., Machado, P., Paquete, L., Whitley, D. (eds.) PPSN 2018. LNCS, vol. 11102, pp. 387–398. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-99259-4_31
Le, N., O’Neill, M., Brabazon, A.: How learning strategies can promote an evolving population in dynamic environments. In: 2019 IEEE Congress on Evolutionary Computation (CEC), June 2019, pp. 2284–2291 (2019)
Le, N., O’Neill, M., Brabazon, A.: Evolutionary consequences of learning strategies in a dynamic rugged landscape. In: Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2019), pp. 812–819. ACM, New York (2019). https://doi.org/10.1145/3321707.3321741
Suzuki, R., Arita, T.: Repeated occurrences of the Baldwin effect can guide evolution on rugged fitness landscapes. In: 2007 IEEE Symposium on Artificial Life. IEEE, April 2007
Keesing, R., Stork, D.G.: Evolution and learning in neural networks: the number and distribution of learning trials affect the rate of evolution. In: NIPS 1990 (1990)
Fernando, C.T., et al.: Meta-learning by the Baldwin effect. CoRR, vol. abs/1806.07917 (2018). http://arxiv.org/abs/1806.07917
Bengio, S., Bengio, Y., Cloutier, J., Gecsei, J.: On the optimization of a synaptic learning rule. In: Levine, D.S., Elsberry, W.R. (eds.) Optimality in Biological and Artificial Networks. Lawrence Erlbaum, New York (1995)
Heyes, C.M.: Social learning in animals: categories and mechanisms. Biol. Rev. 69(2), 207–231 (1994)
Feldman, M.W., Aoki, K., Kumm, J.: Individual versus social learning: evolutionary analysis in a fluctuating environment. Working Papers, Santa Fe Institute (1996)
Enquist, M., Eriksson, K., Ghirlanda, S.: Critical social learning: a solution to Rogers’ paradox of nonadaptive culture. Am. Anthropol. 109(4), 727–734 (2007). https://doi.org/10.1525/aa.2007.109.4.727
Marriott, C., Borg, J.M., Andras, P., Smaldino, P.E.: Social learning and cultural evolution in artificial life. Artif. Life 24(1), 5–9 (2018)
Acerbi, A., Nolfi, S.: Social learning and cultural evolution in embodied and situated agents. In: 2007 IEEE Symposium on Artificial Life. IEEE, April 2007
Le, N.: Evolving self-supervised neural networks: autonomous intelligence from evolved self-teaching. CoRR, vol. abs/1906.08865 (2019). http://arxiv.org/abs/1906.08865
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Le, N., Brabazon, A., O’Neill, M. (2020). Social Learning vs Self-teaching in a Multi-agent Neural Network System. In: Castillo, P.A., Jiménez Laredo, J.L., Fernández de Vega, F. (eds) Applications of Evolutionary Computation. EvoApplications 2020. Lecture Notes in Computer Science(), vol 12104. Springer, Cham. https://doi.org/10.1007/978-3-030-43722-0_23
Download citation
DOI: https://doi.org/10.1007/978-3-030-43722-0_23
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-43721-3
Online ISBN: 978-3-030-43722-0
eBook Packages: Computer ScienceComputer Science (R0)