Abstract
Sorting data into groups and clusters is one of the fundamental tasks of artificially intelligent systems. Classical clustering algorithms rely on heuristic (k-nearest neighbours) or statistical methods (k-means, fuzzy c-means) to derive clusters and these have performed well. Neural networks have also been used in clustering data, but researchers have only recently begun to adopt the strategy of having neural networks directly determine the cluster membership of an input datum. This paper presents a novel strategy, employing NeuroEvolution of Augmenting Topologies to produce an evoltionary neural network capable of directly clustering unlabelled inputs. It establishes the use of cluster validity metrics in a fitness function to train the neural network.
Original language | English |
---|---|
Title of host publication | GECCO '18: Proceedings of the Genetic and Evolutionary Computation Conference Companion |
Number of pages | 162 |
Publisher | Association for Computing Machinery |
Publication date | Jul 2018 |
Pages | 161 |
DOIs | |
Publication status | Published - Jul 2018 |
Event | GECCO '18: The Genetic and Evolutionary Computation Conference - Kyoto, Japan, Kyoto, Japan Duration: 15 Jul 2018 → 19 Jul 2018 http://gecco-2018.sigevo.org/index.html/tiki-index.html |
Conference
Conference | GECCO '18 |
---|---|
Location | Kyoto, Japan |
Country/Territory | Japan |
City | Kyoto |
Period | 15/07/2018 → 19/07/2018 |
Internet address |
Keywords
- Cluster Analysis
- Neural Networks
- NeuroEvolution of Augmenting Topologies
- Unsupervised Learning
- Cluster Validity Metrics