New Methods for Competitive Coevolution

By Christopher D. Rosin, Richard K. Belew. 1997

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Evolutionary Computation

Citations in the corpus

Prioritized Experience Replay (2015, background): “…even stronger emphasis of diversity may be necessary, for example by tracking the age of each transitions and using it to modulate the priority in such a way as to preserve sufﬁ-cient old experience to prevent cycles (related to ‘hall of fame’ ideas in multi-agent literature, Rosin & Belew, 1997).”
Emergent Tool Use From Multi-Agent Autocurricula (2019, background): “[68] Christopher D Rosin and Richard K Belew.”
Human-level performance in 3D multiplayer games with population-based reinforcement learning (2018, methodology): “Our solution is to train a population of P different agents π = (πp)p=1 in parallel that play with each other, introducing diversity amongst players to stabilise training (54).”
Open-ended Learning in Symmetric Zero-sum Games (2019, background): “There is a substantial literature on co-evolution, which has substantial overlap with our setting (Hillis, 1990; Rosin, 1997; Nolfi & Floreano, 1998; Ficici & Pollack, 2000; 2001; de Jong, 2001; Monroy et al., 2006; Schmidt & Lipson, 2008; Popovici et al., 2012).”
A Game-Theoretic Memory Mechanism for Coevolution (2003, methodology): “Examples of this approach are found in Sims [10] (m =1 ,l =1 ), Rosin and Belew [ 9 ] (e.g., m = ∞, l = 20), and Nolfi and Floreano [7] (m = 10, l = 10).”
Evolution of Cartesian Genetic Programs for Development of Learning Neural Architecture (2011, background): “These 502 Evolutionary Computation Volume 19, Number 3 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/EVCO_a_00043 by guest on 30 May 2021 competing solutions create an “arms race” of increasingly better solutions (Dawkins and Krebs, 1979; Rosin and Belew, 1997; Van Valin, 1973).”
Evolving Neural Networks through Augmenting Topologies (2002, background): “However, evolution tends to find the simplest solutions that can win, meaning that strategies oscillate between different idiosyncratic yet uninteresting variations (Darwen, 1996; Rosin and Belew, 1997).”
Evolving a Roving Eye for Go (2004, methodology): “Well-founded coevolution methodologies such as Host-Parasite, Hall of Fame, or Pareto coevolution should be used to develop the most well-rounded players possible [19,20].”
Cooperative Multi-Agent Learning: The State of the Art (2005, background): “[191] C.”
Evolutionary Computation for Reinforcement Learning (2012, background): “Such fitness functions can be implemented using competitive fitness sharing (Rosin and Belew, 1997).”
A survey on optimization metaheuristics (2013, background): “Competitive coevolution has been since successfully applied to game playing strategies [231,210], evolving better pattern recognizers [153], coevolve complex agent behaviors [248], etc.”
Biasing Coevolutionary Search for Optimal Multiagent Behaviors (2006, methodology): “To some extent, the “Hall of Fame” method introduced by [41] for competitive coevolution is also related to biased coevolution; however, that technique samples randomly in the Hall of Fame to increase robustness, while we tend to deterministically select the ideal partners.”
Learning and Evolution (1999, background): “One possible way to force coevolution to produce general strategies is to save and use as competitors all the best individuals of previous generations (see Rosin and Belew, 1997).”
A Competitive-Cooperative Coevolutionary Paradigm for Dynamic Multiobjective Optimization (2009, methodology): “When applying this idea into optimization [1], [51], there are usually two subpopulations and an inverse ﬁtness interaction exists between the two subpopulations.”
Opposition based learning: A literature review (2017, background): “In the HOF strategy [227], each individual is competing against every archived best individual from previous generations.”
Competitive Coevolution through Evolutionary Complexification (2011, methodology): “Evolution tends to find the simplest solutions that can win, meaning that strategies can switch back and forth between dierent idiosyncratic yet uninteresting variations (Darwen, 1996; Floreano & Nolfi, 1997; Rosin & Belew, 1997 ).”
Modelling brain emergent behaviours through coevolution of neural agents (2006, background): “Most of the coevolutionary approaches presented in the literature can be classified as competitive [36], or collaborative [33].”