Loading…
NECEC 2017 has ended
View analytic
Wednesday, November 15 • 14:00 - 14:20
A New Mapless Clustering Algorithm

Sign up or log in to save this to your schedule and see who's attending!

Self-organizing maps have been used for several decades as an artificial neural network model for dimension reduction of high-dimensional data for the purposes of data clustering and visualization. In this paper, we propose a new data clustering method inspired from the self-organizing map. Unlike the self-organizing map, we propose a pre-clustering phase for our method that uses Euclidean distance and cosine similarity. Once data are pre-clustered, this method uses an unsupervised learning phase (training phase), which is similar to the training phase for self-organizing maps. Unlike traditional self-organizing maps, the proposed algorithm does not use a map at all, mitigating the issues associated with its use. The method is also highly-parallelizable which allows for a GPU implementation. Results demonstrating the excellent performance of the proposed clustering algorithm are included in this paper.


Wednesday November 15, 2017 14:00 - 14:20
Conception Bay South 180 Portugal Cove Road, St. John's, NL, Canada

Attendees (1)