Ieee Transactions on Evolutionary Computation on the Performance of Indirect Encoding across the Continuum of Regularity

By Jeff Clune, Kenneth O. Stanley, Robert T. Pennock, Charles Ofria, Stanley

How researchers see this paper

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

Citations in the corpus

Deep learning in neural networks: An overview (2014, neutral): “Instead of evolving large NNs directly (Section 6.6), one can sometimes greatly reduce the search space by evolving compact encodings of NNs, e.g., through Lindenmeyer Systems (Jacob, Lindenmayer, & Rozenberg, 1994; Lindenmayer, 1968), graph rewriting (Kitano, 1990), Cellular Encoding (Gruau, Whitle”
Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternative for Training Deep Neural Networks for Reinforcement Learning (2017, background): “, 2003; Haupt & Haupt, 2004), including crossover (Holland, 1992; Deb & Myburgh, 2016), indirect encoding (Stanley, 2007; Stanley et al., 2009; Clune et al., 2011), and encouraging quality diversity (Mouret & Clune, 2015; Pugh et al.”
Paired Open-Ended Trailblazer (POET): Endlessly Generating Increasingly Complex and Diverse Learning Environments and Their Solutions (2019, background): “Work evolving robot controllers with HyperNEAT has shown it can produce neural networks with regular geometric patterns in neural connectivity that exploit regularities in the environment to produce regular behaviors that look natural and improve performance [68, 69, 69, 70].”
On the Relationships between Generative Encodings, Regularity, and Learning Abilities when Evolving Plastic Artificial Neural Networks (2013, background): “Inﬂuenced by this concept, many researchers proposed artiﬁcial developmental systems with diverse inspirations including chemical gradients [11,12], gene regulatory networks [13,14], cell divisions [15], computational neuro-science models [16] and L-systems [9].”
From evolutionary computation to the evolution of things (2015, background): “Such generative and developmental representations allow the reuse of code which helps scale up the complexity of artificially-evolved phenotypes, for instance, in evolutionary robotics, artificial life, and morphogenetic engineering 81,82,83,84,85 .”
Guided self-organization in indirectly encoded and evolving topographic maps (2014, unknown): “Because of this insight, HyperNEAT is able to evolve high-dimensional ANNs with regularities appropriate to the problem domain [4, 8, 9, 29], which is in the case of this paper a self-organizing topographic map.”
Neuroevolution of self-interpretable agents (2020, methodology): “t on the choice of our indirect encoding method. arXiv:2003.08165v1 [cs.NE] 18 Mar 2020 Yujin Tang, Duong Nguyen, and David Ha For instance, HyperNEAT [90] has been successful at robotic gait control [14, 15, 75], suggesting CPPNs [89] to be effective at representing modular and symmetric properties”
AI-GAs: AI-generating algorithms, an alternate paradigm for producing general artificial intelligence (2019, background): “That includes research into encodings (aka representations) [168, 171, 170, 203, 55], search operators, and the production of architectures that are regular [171, 27, 69], modular [28, 69], and hierarchical [109].”
Evolvability Search: Directly Selecting for Evolvability in order to Study and Produce It (2016, background): “Accordingly, researchers have proposed many approaches for encouraging evolvability [5,6,12,16,18,20,27,30].”
Deep neural networks are easily fooled: High confidence predictions for unrecognizable images (2014, methodology): “We test EAs with two different encodings [29, 5], meaning how an image is represented as a genome.”
Critical factors in the performance of hyperNEAT (2013, methodology): “Finally, we connect these results to previous experiments showing that HyperNEAT’s performance decreases as irregularity in the required solution increases [7].”
Neuromodulation Improves the Evolution of Forward Models (2016, background): “Because we are interested in learning not only how to produce effective forward models for robotic control, but also to learn about the evolution of forward models in animals, the forward models in this paper are produced by evolving neural networks, a field called neuroevolution [5, 6, 10, 12, 27].”
Lifted Relational Neural Networks: Efficient Learning of Latent Relational Structures (2018, background): “Another line of work are convolutional neural networks (LeCun, Bottou, Bengio, & Haﬀner, 1998) and techniques of indirect encoding (Clune, Stanley, Pennock, & Ofria, 2011), exploiting patterns and regularities in neural connections to create more compressed representations of large neural networks.”
Artificial Evolution of Plastic Neural Networks: A Few Key Concepts (2014, background): “other generative encodings such as HyperNEAT [9], [30] could also possess a low TLA-level.”