Neuroevolution: Evolving Novel Neural Network Architectures with Kenneth Stanley - TWiML Talk #94
Published:Jan 11, 2018 01:08
•1 min read
•Practical AI
Analysis
This article discusses neuroevolution, a method of evolving neural network architectures using genetic algorithms. It features an interview with Kenneth Stanley, a leading researcher in this field. The conversation covers Stanley's work, including the Neuroevolution of Augmenting Topologies (NEAT) paper, HyperNEAT, and novelty search. The article highlights the potential of neuroevolution to create more complex and human-like neural networks, as well as approaches that prioritize novel behaviors over predefined objectives. The discussion also touches upon the relationship between biology and computation, and Stanley's other projects.
Key Takeaways
- •Neuroevolution uses genetic algorithms to evolve neural network architectures.
- •NEAT, HyperNEAT, and novelty search are key approaches discussed.
- •The research explores creating more complex and human-like networks and novel behaviors.
Reference
“The article doesn't contain a specific quote to extract.”