Technology Category
Machine Learning: Evolutionary LearningSearch and Optimization: Evolutionary ComputationCognitive Modeling & Cognitive Systems: Neural Spike Coding
Introducing Echo Networks for Computational Neuroevolution
🤖 AI Summary
This work addresses the challenge of efficiently detecting and classifying events in discrete-time signals on ultra-edge devices, where existing lightweight neural networks struggle due to the absence of structured mutation and recombination mechanisms. To overcome this limitation, the authors propose Echo Networks—a hierarchical-free recurrent architecture composed solely of a connectivity matrix, allowing any neuron to serve as input or output and encoding the entire network genome into a single matrix. Building upon this unified representation, they introduce mutation and recombination operators grounded in matrix operations and decompositions, which substantially enhance the systematicity and flexibility of evolutionary processes. Evaluated on electrocardiogram signal classification, Echo Networks demonstrate remarkable evolutionary efficiency and competitive classification performance despite their minimal structural complexity.
Technology Category
Application Category
📝 Abstract
For applications on the extreme edge, minimal networks of only a few dozen artificial neurons for event detection and classification in discrete time signals would be highly desirable. Feed-forward networks, RNNs, and CNNs evolved through evolutionary algorithms can all be successful in this respect but pose the problem of allowing little systematicity in mutation and recombination if the standard direct genetic encoding of the weights is used (as for instance in the classic NEAT algorithm). We therefore introduce Echo Networks, a type of recurrent network that consists of the connection matrix only, with the source neurons of the synapses represented as rows, destination neurons as columns and weights as entries. There are no layers, and connections between neurons can be bidirectional but are technically all recurrent. Input and output can be arbitrarily assigned to any of the neurons and only use an additional (optional) function in their computational path, e.g., a sigmoid to obtain a binary classification output. We evaluated Echo Networks successfully on the classification of electrocardiography signals but see the most promising potential in their genome representation as a single matrix, allowing matrix computations and factorisations as mutation and recombination operators.
Problem
Research questions and friction points this paper is trying to address.
neuroevolution
genetic encoding
recurrent networks
mutation operators
edge computing
Innovation
Methods, ideas, or system contributions that make the work stand out.
Echo Networks
computational neuroevolution
matrix-based encoding
recurrent neural networks
matrix factorization
🔎 Similar Papers
No similar papers found.
