Evolving artificial neural networks

By Xin Yao. 1999

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

Real-Time Federated Evolutionary Neural Architecture Search (2020, positive): “Different from the previous neuroevolution techniques [35], [36] that aim to optimize both the weights and architecture of neural networks, EA based NAS only optimizes the model architecture itself, and the model parameters are trained using conventional gradient descent methods [10]–[13].”
Genetic CNN (2017, methodology): “We also note that some previous work applied the genetic algorithm to learning the structure [35][1] or weights [41][6] of artificial neural networks, but our work aims at learning the architecture of modern CNNs, which is not studied in prior researches.”
Diffusion-based neuromodulation can eliminate catastrophic forgetting in simple neural networks (2017, methodology): “EAs are search algorithms based on Darwinian evolution, and can search through various ANN configurations for the right weights that allow an ANN to solve a problem [9,22].”
AI-GAs: AI-generating algorithms, an alternate paradigm for producing general artificial intelligence (2019, background): “However, there is also a long history in AI of creating algorithms to search for high-performing neural network architectures [90, 169, 55, 198, 59].”
A Comprehensive Analysis of Deep Regression (2018, methodology): “[54], [55] to second-order methods [53], but their use is much less widespread, so they have been excluded from”
Evolving deep unsupervised convolutional networks for vision-based reinforcement learning (2014, background): “neuroevolution; [24]) except that during evaluation the candidate networks do not receive observations directly from the environment.”
Supervised learning of fuzzy ARTMAP neural networks through particle swarm optimization (2007, background): “According to Yao [57], the concept of evolution has been introduced into neural networks at three different levels: connection weights, architectures, and learning rules.”
Evolutionary computation: a unified approach (2018, background): “Consequently, a number of hybrid EA-ANN systems have been developed to make weight learning more robust (Yao, 1999).”
Time-series prediction using a local linear wavelet neural network (2006, background): “Evolutionary algorithms including PSO are global search algorithms but suffer from fine-tuning inefficiency [ 32 ].”
Hybrid Genetic Algorithms: A Review (2006, background): “For a comprehensive review of evolving neural networks the reader can refer to [91].”
Generative NeuroEvolution for Deep Learning (2013, background): “Neuro-evolution (NE; [3]) methods train artificial neural networks (ANNs) through evolutionary algorithms.”
Block-Based Neural Networks for Personalized ECG Signal Classification (2007, background): “Evolvable NNs [ 18 ] change the network structure and internal configurations as well as the parameters to cope with dynamic operating environments.”
Evolution strategies – A comprehensive introduction (2002, unknown): “In such applications, the evolution of the structure is usually performed by the outer ES, while the weights are evolved by the inner ES (see, e.g., Yao, 1999).”