{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,17]],"date-time":"2026-02-17T02:28:11Z","timestamp":1771295291056,"version":"3.50.1"},"reference-count":49,"publisher":"MIT Press - Journals","issue":"1","license":[{"start":{"date-parts":[[2021,2,5]],"date-time":"2021-02-05T00:00:00Z","timestamp":1612483200000},"content-version":"vor","delay-in-days":401,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"content-domain":{"domain":["direct.mit.edu"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2020,3,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:p>The structure and performance of neural networks are intimately connected, and by use of evolutionary algorithms, neural network structures optimally adapted to a given task can be explored. Guiding such neuroevolution with additional objectives related to network structure has been shown to improve performance in some cases, especially when modular neural networks are beneficial. However, apart from objectives aiming to make networks more modular, such structural objectives have not been widely explored. We propose two new structural objectives and test their ability to guide evolving neural networks on two problems which can benefit from decomposition into subtasks. The first structural objective guides evolution to align neural networks with a user-recommended decomposition pattern. Intuitively, this should be a powerful guiding target for problems where human users can easily identify a structure. The second structural objective guides evolution towards a population with a high diversity in decomposition patterns. This results in exploration of many different ways to decompose a problem, allowing evolution to find good decompositions faster. Tests on our target problems reveal that both methods perform well on a problem with a very clear and decomposable structure. However, on a problem where the optimal decomposition is less obvious, the structural diversity objective is found to outcompete other structural objectives\u2014and this technique can even increase performance on problems without any decomposable structure at all.<\/jats:p>","DOI":"10.1162\/evco_a_00250","type":"journal-article","created":{"date-parts":[[2019,2,15]],"date-time":"2019-02-15T16:43:49Z","timestamp":1550249029000},"page":"115-140","update-policy":"https:\/\/doi.org\/10.1162\/mitpressjournals.corrections.policy","source":"Crossref","is-referenced-by-count":11,"title":["Guiding Neuroevolution with Structural Objectives"],"prefix":"10.1162","volume":"28","author":[{"given":"Kai Olav","family":"Ellefsen","sequence":"first","affiliation":[{"name":"Department of Informatics, University of Oslo, Norway kaiolae@ifi.uio.no"}]},{"given":"Joost","family":"Huizinga","sequence":"additional","affiliation":[{"name":"Uber AI Labs, USA jhuizinga@uber.com"}]},{"given":"Jim","family":"Torresen","sequence":"additional","affiliation":[{"name":"Department of Informatics and RITMO, University of Oslo, Norway jimtoer@ifi.uio.no"}]}],"member":"281","published-online":{"date-parts":[[2020,3,1]]},"reference":[{"key":"2022050416562523200_B1","doi-asserted-by":"crossref","unstructured":"Alon, U\n          . (2006). An introduction to systems biology: Design\n            principles of biological circuits. Boca Raton,\n            FL: Chapman &\n          Hall\/CRC.","DOI":"10.1201\/9781420011432"},{"key":"2022050416562523200_B2","doi-asserted-by":"crossref","unstructured":"Beer, R. D., and\n                Gallagher, J.\n              C. (1992).\n            Evolving dynamical neural networks for adaptive behavior.\n            Adaptive Behavior,\n            1(1):91.","DOI":"10.1177\/105971239200100105"},{"key":"2022050416562523200_B3","doi-asserted-by":"crossref","unstructured":"Brandes, U.,\n                Delling,\n            D., Gaertler,\n                M., G\u00f6rke,\n                R., Hoefer,\n                M.,\n                Nikoloski,\n              Z., and\n                Wagner,\n            D. (2008).\n            On modularity clustering. IEEE Transactions on\n            Knowledge and Data Engineering,\n            20(2):172\u2013188.","DOI":"10.1109\/TKDE.2007.190689"},{"key":"2022050416562523200_B4","doi-asserted-by":"crossref","unstructured":"Brooks, R. A\n          . (1986). A robust layered control system for a\n            mobile robot. IEEE Journal on Robotics and Automation,\n            2(1):14\u201323.","DOI":"10.1109\/JRA.1986.1087032"},{"key":"2022050416562523200_B5","doi-asserted-by":"crossref","unstructured":"Cardamone, L.,\n                Loiacono,\n              D., and\n              Lanzi, P.\n            L. (2009).\n            Evolving competitive car controllers for racing games with\n            neuroevolution. In Proceedings of the 11th Annual Conference on\n            Genetic and Evolutionary Computation (GECCO), pp.\n            1179\u20131186.","DOI":"10.1145\/1569901.1570060"},{"key":"2022050416562523200_B6","doi-asserted-by":"crossref","unstructured":"Clune, J.,\n                Beckmann, B.\n              E., McKinley,\n                P. K., and\n                Ofria,\n            C. (2010).\n            Investigating whether HyperNEAT produces modular neural\n            networks. In Proceedings of the Genetic and Evolutionary\n            Computation Conference, pp.\n          635\u2013642.","DOI":"10.1145\/1830483.1830598"},{"key":"2022050416562523200_B7","doi-asserted-by":"crossref","unstructured":"Clune, J.,\n                Mouret,\n              J.-B., and\n                Lipson,\n            H. (2013).\n            The evolutionary origins of modularity. In\n            Proceedings of the Royal Society of London. Series B: Biological\n            Sciences,\n          280(1755):20122863.","DOI":"10.1098\/rspb.2012.2863"},{"key":"2022050416562523200_B8","unstructured":"Conti, E.,\n                Madhavan,\n              V., Petroski\n                Such, F.,\n                Lehman, J.,\n                Stanley,\n            K., and Clune,\n                J. (2018).\n            Improving exploration in evolution strategies for deep reinforcement\n            learning via a population of novelty-seeking agents. In S.Bengio,\n                H.Wallach,\n                H.Larochelle,\n                K.Grauman,\n                N.Cesa-Bianchi,\n            and\n            R.Garnett\n          (Eds.), Advances in neural information processing\n            systems31, pp. 5027\u20135038.\n            Red Hook, NY: Curran Associates,\n            Inc."},{"key":"2022050416562523200_B9","doi-asserted-by":"crossref","unstructured":"Deb, K.,\n                Pratap, A.,\n                Agarwal,\n            S., and\n                Meyarivan,\n              T. (2002).\n            A fast and elitist multiobjective genetic algorithm:\n            NSGA-II. IEEE Transactions on Evolutionary Computation,\n            6(2):182\u2013197.","DOI":"10.1109\/4235.996017"},{"key":"2022050416562523200_B10","doi-asserted-by":"crossref","unstructured":"Eiben, A. E., and\n                Smith, J.\n              E. (2003).\n              Introduction to evolutionary\n          computing. New York:\n            Springer.","DOI":"10.1007\/978-3-662-05094-1"},{"key":"2022050416562523200_B11","doi-asserted-by":"crossref","unstructured":"Ellefsen, K. O.,\n                Mouret,\n              J.-B., and\n                Clune,\n            J. (2015).\n            Neural modularity helps organisms evolve to learn new skills without\n            forgetting old skills. PLOS Computational Biology,\n            11(4):e1004128.","DOI":"10.1371\/journal.pcbi.1004128"},{"key":"2022050416562523200_B12","doi-asserted-by":"crossref","unstructured":"Ellefsen, K. O., and\n                Torresen,J.\n            (2017). Evolving neural networks with multiple internal\n            models. In Proceedings of the European Conference on Artificial\n            Life, 1414:138\u2013145.","DOI":"10.7551\/ecal_a_025"},{"key":"2022050416562523200_B13","doi-asserted-by":"crossref","unstructured":"Espinosa-Soto, C., and\n                Wagner,\n            A. (2010).\n            Specialization can drive the evolution of modularity.\n            PLoS Computational Biology,\n            6(3):e1000719.","DOI":"10.1371\/journal.pcbi.1000719"},{"key":"2022050416562523200_B14","doi-asserted-by":"crossref","unstructured":"Fortunato, S.\n           (2010). Community detection in\n            graphs. Physics Reports,\n            486:75\u2013174.","DOI":"10.1016\/j.physrep.2009.11.002"},{"key":"2022050416562523200_B15","unstructured":"Goldberg, D. E., and\n                Richardson,\n              J. (1987).\n            Genetic algorithms with sharing for multimodal function\n            optimization. In Genetic Algorithms and Their Applications:\n            Proceedings of the Second International Conference on Genetic Algorithms, pp.\n            41\u201349."},{"key":"2022050416562523200_B16","doi-asserted-by":"crossref","unstructured":"Gruau, F\n          . (1994). Automatic definition of modular\n            neural networks. Adaptive Behavior,\n            3(2):151\u2013183.","DOI":"10.1177\/105971239400300202"},{"key":"2022050416562523200_B17","doi-asserted-by":"crossref","unstructured":"H\u00f8verstad, B. A.\n           (2011). Noise and the evolution of neural\n            network modularity. Artificial Life,\n            17(1):33\u201350.","DOI":"10.1162\/artl_a_00016"},{"key":"2022050416562523200_B18","doi-asserted-by":"crossref","unstructured":"Huizinga, J.,\n                Mouret,\n              J.-B., and\n                Clune,\n            J. (2014).\n            Evolving neural networks that are both modular and regular: HyperNeat\n            plus the connection cost technique. In Proceedings of the\n            Genetic and Evolutionary Computation Conference, pp.\n            697\u2013704.","DOI":"10.1145\/2576768.2598232"},{"key":"2022050416562523200_B19","doi-asserted-by":"crossref","unstructured":"Huizinga, J.,\n                Mouret,\n              J.-B., and\n                Clune,\n            J. (2016).\n            Does aligning phenotypic and genotypic modularity improve the evolution\n            of neural networks? In Proceedings of the 2016 on Genetic and\n            Evolutionary Computation Conference (GECCO), pp.\n            125\u2013132.","DOI":"10.1145\/2908812.2908836"},{"key":"2022050416562523200_B20","doi-asserted-by":"crossref","unstructured":"Kashtan, N., and\n                Alon,\n            U. (2005).\n            Spontaneous evolution of modularity and network motifs.\n            Proceedings of the National Academy of Sciences,\n            102(39):13773\u201313778.","DOI":"10.1073\/pnas.0503610102"},{"key":"2022050416562523200_B21","doi-asserted-by":"crossref","unstructured":"LeCun, Y.,\n                Bengio, Y.,\n            and Hinton,\n            G. (2015).\n            Deep learning. Nature,\n            521(7553):436\u2013444.","DOI":"10.1038\/nature14539"},{"key":"2022050416562523200_B22","doi-asserted-by":"crossref","unstructured":"Leicht, E. A., and\n                Newman, M. E.\n              J. (2008).\n            Community structure in directed networks.\n            Physical Review Letters,\n          118703\u2013118707.","DOI":"10.1103\/PhysRevLett.100.118703"},{"key":"2022050416562523200_B23","doi-asserted-by":"crossref","unstructured":"Mengistu, H., and\n                Clune,\n            J. (2016).\n            The evolutionary origins of hierarchy. PLOS\n            Computational Biology, 12(6).","DOI":"10.1371\/journal.pcbi.1004829"},{"key":"2022050416562523200_B24","doi-asserted-by":"crossref","unstructured":"Mnih, V.,\n                Kavukcuoglu,\n              K., Silver,\n                D., Rusu,\n                A. A.,\n                Veness, J.,\n                Bellemare, M.\n              G., Graves,\n                A.et al, et al.\n            (2015). Human-level control through deep reinforcement\n            learning. Nature,\n            518(7540):529.","DOI":"10.1038\/nature14236"},{"key":"2022050416562523200_B25","doi-asserted-by":"crossref","unstructured":"Mountcastle, V. B.\n           (1997). The columnar organization of the\n            neocortex. Brain,\n            120(4):701\u2013722.","DOI":"10.1093\/brain\/120.4.701"},{"key":"2022050416562523200_B26","doi-asserted-by":"crossref","unstructured":"Mouret, J.-B., and\n                Doncieux,\n            S. (2008).\n            MENNAG: A modular, regular and hierarchical encoding for neural-networks\n            based on attribute grammars. Evolutionary Intelligence,\n            1(3):187\u2013207.","DOI":"10.1007\/s12065-008-0015-7"},{"key":"2022050416562523200_B27","doi-asserted-by":"crossref","unstructured":"Mouret, J.-B., and\n                Doncieux,\n            S. (2009).\n            Using behavioral exploration objectives to solve deceptive problems in\n            neuro-evolution. In Proceedings of the 11th Annual Conference on\n            Genetic and Evolutionary Computation, pp.\n          627\u2013634.","DOI":"10.1145\/1569901.1569988"},{"key":"2022050416562523200_B28","unstructured":"Mouret, J.-B., and\n                Doncieux,\n              S. (2010).\n            SFERES v2: Evolvin' in the multi-core world. In\n            IEEE Congress on Evolutionary Computation2, pp.\n            4079\u20134086."},{"key":"2022050416562523200_B29","doi-asserted-by":"crossref","unstructured":"Mouret, J.-B., and\n                Doncieux,\n            S. (2012).\n            Encouraging behavioral diversity in evolutionary robotics: An empirical\n            study. Evolutionary Computation,\n            1(20):91\u2013113.","DOI":"10.1162\/EVCO_a_00048"},{"key":"2022050416562523200_B30","doi-asserted-by":"crossref","unstructured":"Newman, M. E\n          . (2006a). Finding community structure in\n            networks using the eigenvectors of matrices. Physical Review\n            E,\n          74(3):036104.","DOI":"10.1103\/PhysRevE.74.036104"},{"key":"2022050416562523200_B31","doi-asserted-by":"crossref","unstructured":"Newman, M. E. J\n          . (2006b). Modularity and community structure\n            in networks. Proceedings of the National Academy of Sciences of\n            the United States of America,\n            103(23):8577\u20138582.","DOI":"10.1073\/pnas.0601602103"},{"key":"2022050416562523200_B32","doi-asserted-by":"crossref","unstructured":"Poulsen, A. P.,\n                Thorhauge,\n              M., Funch,\n                M. H., and\n                Risi,\n            S. (2017).\n            DLNE: A hybridization of deep learning and neuroevolution for visual\n            control. In IEEE Conference on Computational Intelligence and\n            Games, pp. 256\u2013263.","DOI":"10.1109\/CIG.2017.8080444"},{"key":"2022050416562523200_B33","doi-asserted-by":"crossref","unstructured":"Pugh, J. K., and\n                Stanley, K.\n              O. (2013).\n            Evolving multimodal controllers with hyperneat. In\n            Proceedings of the 15th Annual Conference on Genetic and Evolutionary\n            Computation, pp. 735\u2013742.","DOI":"10.1145\/2463372.2463459"},{"key":"2022050416562523200_B34","doi-asserted-by":"crossref","unstructured":"Real, E.,\n                Aggarwal,\n              A., Huang,\n                Y., and Le,\n                Q. V. (2018).\n            Regularized evolution for image classifier architecture\n            search. Retrieved from\n          arXiv:1802.01548.","DOI":"10.1609\/aaai.v33i01.33014780"},{"key":"2022050416562523200_B35","doi-asserted-by":"crossref","unstructured":"Risi, S.,\n                Vanderbleek, S.\n              D., Hughes,\n                C. E., and\n                Stanley, K.\n              O. (2009). How\n            novelty search escapes the deceptive trap of learning to learn. In\n            Proceedings of the 11th Annual Conference on Genetic and Evolutionary\n            Computation, pp. 153\u2013160.","DOI":"10.1145\/1569901.1569923"},{"key":"2022050416562523200_B36","unstructured":"Salimans, T.,\n                Ho, J.,\n                Chen, X.,\n                Sidor, S.,\n            and Sutskever,\n              I. (2017).\n            Evolution strategies as a scalable alternative to reinforcement\n            learning. Retrieved from\n          arXiv:1702.03864."},{"key":"2022050416562523200_B37","doi-asserted-by":"crossref","unstructured":"Schrum, J., and\n                Miikkulainen,\n              R. (2012).\n            Evolving multimodal networks for multitask games.\n            IEEE Transactions on Computational Intelligence and AI in Games,\n            4(2):94\u2013111.","DOI":"10.1109\/TCIAIG.2012.2193399"},{"key":"2022050416562523200_B38","doi-asserted-by":"crossref","unstructured":"Schrum, J., and\n                Miikkulainen,\n              R. (2014).\n            Evolving multimodal behavior with modular neural networks in Ms.\n            Pac-Man. In Proceedings of the 2014 Conference on Genetic and\n            Evolutionary Computation (GECCO), pp.\n          325\u2013332.","DOI":"10.1145\/2576768.2598234"},{"key":"2022050416562523200_B39","doi-asserted-by":"crossref","unstructured":"Schrum, J., and\n                Miikkulainen,\n              R. (2016a).\n            Discovering multimodal behavior in Ms. Pac-Man through evolution of\n            modular neural networks. IEEE Transactions on Computational\n            Intelligence and AI in Games,\n            8(1):67\u201381.","DOI":"10.1109\/TCIAIG.2015.2390615"},{"key":"2022050416562523200_B40","doi-asserted-by":"crossref","unstructured":"Schrum, J., and\n                Miikkulainen,\n              R. (2016b).\n            Solving multiple isolated, interleaved, and blended tasks through modular\n            neuroevolution. Evolutionary Computation,\n            24(3):459\u2013490.","DOI":"10.1162\/EVCO_a_00181"},{"key":"2022050416562523200_B41","unstructured":"Simonyan, K., and\n                Zisserman,\n              A. (2014).\n            Very deep convolutional networks for large-scale image\n            recognition. Retrieved from CoRR,\n          abs\/1409.1556."},{"key":"2022050416562523200_B42","doi-asserted-by":"crossref","unstructured":"Stanley, K. O\n          . (2007). Compositional pattern producing\n            networks: A novel abstraction of development. Genetic\n            Programming and Evolvable Machines,\n            8(2):131\u2013162.","DOI":"10.1007\/s10710-007-9028-8"},{"key":"2022050416562523200_B43","doi-asserted-by":"crossref","unstructured":"Stanley, K. O.,\n                D'Ambrosio, D.\n              B., and Gauci,\n                J. (2009).\n            A hypercube-based encoding for evolving large-scale neural\n            networks. Artificial Life,\n            15(2):185\u2013212.","DOI":"10.1162\/artl.2009.15.2.15202"},{"key":"2022050416562523200_B44","doi-asserted-by":"crossref","unstructured":"Stanley, K. O., and\n                Miikkulainen,\n              R. (2002).\n            Evolving neural network through augmenting topologies.\n            Evolutionary Computation,\n            10(2):99\u2013127.","DOI":"10.1162\/106365602320169811"},{"key":"2022050416562523200_B45","unstructured":"Such, F. P.,\n                Madhavan,\n              V., Conti,\n                E., Lehman,\n                J.,\n              Stanley, K.\n              O., and Clune,\n                J. (2017).\n            Deep neuroevolution: Genetic algorithms are a competitive alternative for\n            training deep neural networks for reinforcement learning.\n            Retrieved from arXiv:1712.06567."},{"key":"2022050416562523200_B46","unstructured":"Togelius, J\n          . (2004). Evolution of a subsumption\n            architecture neurocontroller. Journal of Intelligent and Fuzzy\n            Systems,\n            15(1):15\u201320."},{"key":"2022050416562523200_B47","doi-asserted-by":"crossref","unstructured":"Velez, R., and\n                Clune,\n            J. (2017).\n            Diffusion-based neuromodulation can eliminate catastrophic forgetting in\n            simple neural networks. PLoS ONE,\n            12(11).","DOI":"10.1371\/journal.pone.0187736"},{"key":"2022050416562523200_B48","doi-asserted-by":"crossref","unstructured":"Verbancsics, P., and\n                Stanley, K.\n              O. (2011).\n            Constraining connectivity to encourage modularity in\n            HyperNEAT. In Proceedings of the Genetic and Evolutionary\n            Computation Conference (GECCO), pp.\n          1483\u20131490.","DOI":"10.1145\/2001576.2001776"},{"key":"2022050416562523200_B49","doi-asserted-by":"crossref","unstructured":"Yao, X\n          . (1999). Evolving artificial neural\n            networks. Proceedings of the IEEE,\n            87(9):1423\u20131447.","DOI":"10.1109\/5.784219"}],"container-title":["Evolutionary Computation"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/direct.mit.edu\/evco\/article-pdf\/28\/1\/115\/2020382\/evco_a_00250.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/direct.mit.edu\/evco\/article-pdf\/28\/1\/115\/2020382\/evco_a_00250.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,5,4]],"date-time":"2022-05-04T16:57:04Z","timestamp":1651683424000},"score":1,"resource":{"primary":{"URL":"https:\/\/direct.mit.edu\/evco\/article\/28\/1\/115\/94988\/Guiding-Neuroevolution-with-Structural-Objectives"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020]]},"references-count":49,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2020,3,1]]},"published-print":{"date-parts":[[2020,3,1]]}},"URL":"https:\/\/doi.org\/10.1162\/evco_a_00250","relation":{},"ISSN":["1530-9304"],"issn-type":[{"value":"1530-9304","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2020]]},"published":{"date-parts":[[2020]]}}}