China 
Africa 
Explore chapters and articles related to this topic
Hybrid Cartesian Genetic Programming Algorithms: A Review
Published in Siddhartha Bhattacharyya, Václav Snášel, Indrajit Pan, Debashis De, Hybrid Computational Intelligence, 2019
Johnathan Melo Neto, Heder S. Bernardino, Helio J.C. Barbosa
A more recent optimization approach for ANNs is neuroevolution, which is a sub-area of machine learning that executes EAs to train ANNs [77]. In 2010, Khan et al. [39] introduced a neuroevolutionary method called CGPANN in order to use CGP to the training of ANNs. That proposal can optimize the topology and the weights using (1 + λ)-ES. CGPANN extends the traditional CGP by adding connection weight genes for each connection gene, as illustrated in Figure 2.3. It can be noted that the graph representation of CGPANN is suitable to describe ANNs. However, as CGP is designed to handle discrete problems [52], the CGPANN applied in the ANNs weights can lead to results below the ideal. To circumvent this issue, a better-suited search technique may be used to assist CGP in finding more accurate weights for the ANNs.
A Case for Personalized Non-Player Character Companion Design
Published in International Journal of Human–Computer Interaction, 2023
Emma J. Pretty, Haytham M. Fayek, Fabio Zambetta
Recent advances in AI have opened up many more opportunities for exploring adaptive games, and there are many areas of game design that are yet to be investigated. Some examples of such techniques are Procedural Content Generation (PCG), whereby algorithmic procedures are used to automatically create game content using models of player experience or performance (Togelius et al., 2013; Yannakakis & Togelius, 2011). Other methods include Reinforcement Learning (RL) in which a computer agent’s actions are reinforced based on what decisions it makes, thus over time, learning optimal actions for certain situations (Shao et al., 2019) and Neuroevolution of Augmenting Topologies (NEAT; see Papavasileiou et al. (2021) for a review), which uses evolutionary algorithms to evolve artificial neural networks from RL tasks.