The Evolution of Learning: An Experiment in Genetic Connectionism

By David J. Chalmers. 1991

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

AutoML-Zero: Evolving Machine Learning Algorithms From Scratch (2020, methodology): “An important aspect of an ML algorithm is the optimization of its weights, which has been tackled by AutoML in the form of numerically discovered optimizers (Chalmers, 1991; Andrychowicz et al., 2016; Vanschoren, 2019).”
Born to Learn: the Inspiration, Progress, and Future of Evolved Plastic Artificial Neural Networks (2017, methodology): “s are also among the ﬁrst to include a modulatory term in the learning rules. The optimization was shown to improve the performance in those different tasks with respect to manual parameter settings. Chalmers (1990) evolved a learning rule that applied to every connection and had a teaching signal. ”
Meta-Learning: A Survey (2018, neutral): “(1995) replace backpropagation with simple biologically-inspired parametric rules (or evolved rules (Chalmers, 1991)) to update the synaptic weights.”
On the Relationships between Generative Encodings, Regularity, and Learning Abilities when Evolving Plastic Artificial Neural Networks (2013, background): “the GLA score grows linearly with the size of the evolutionary training set, which is consistent with previous results [28], and the GLA scores obtained with small values of |E| are statistically different from those obtained with larger values (e.”
Meta learning evolutionary artificial neural networks (2004, background): “tive adjustment of BP algorithm's parameters, such as the learning rate and momentum, through evolution could be considered as the first attempt of the evolution of learning rules [40]. Chalmers [23] defined the form of learning rules as a linear function of four local variables and their six p”
Evolving plastic neural networks with novelty search (2010, result): “Interestingly, Chalmers (1990) evolved a global learning rule (i.e., a rule that applies to every connection) and discovered that the evolved rule was similar to the well-known delta rule used in backpropagation.”
Sequential Behavior and Learning in Evolved Dynamical Neural Networks (1994, background): “Finally, there has been a small amount of work on the evolution of learning (Chalmers, 1991; Miller and Todd, 1991) and the relationship between learning and evolution (Hinton and Nowlan, 1987; Belew, McInerney, and Schraudolph, 1990).”
Meta learning evolutionary arti ! cial neural networks (2004, background): “Chalmers [23] de!ned the form of learning rules as a linear function of four local variables and their six pair wise products [11,33].”
Learning to Learn Using Gradient Descent (2001, background): “Several researchers have suggested meta-1 learning systems based on neural networks and used genetic algorithms to adjust the subordinate learning algorithm [2,10,19].”
On the search for new learning rules for ANNs (1995, methodology): “Chalmers proposed in [1] a method to nd new learning rules using evolution mechanisms such as genetic algorithms.”
Evolution of Reinforcement Learning in Uncertain Environments: A Simple Explanation for Complex Foraging Behaviors (2002, background): “A relatively small amount of research has been devoted to the evolution of the learning process itself, most of which was constrained to choosing the appropriate learning rule from a limited set of predefined rules (Baxter, 1992; Chalmers, 1990; Floreano & Mondada, 1996).”
Evolving Plastic Neural Networks for Online Learning: Review and Future Directions (2012, methodology): “Fitness evaluation typically took the form of a training phase during which sets of exemplars are presented to the network, and then an evaluation phase where the network is presented all the input vectors from the training set and the fitness is the percentage of correct corresponding output vector”
Use of genetic programming for the search of a new learning rule for neural networks (1994, methodology): “We proposed in [3] (see also [4] for a similar approach) a method to nd new learning rules using standard optimization techniques.”
Metaheuristic design of feedforward neural networks: A review of two decades of research (2017, background): “However, formulating BP parameter such as learning rule, which is a dynamic concept, into a static chromosome is disadvantageous [253].”