In this paper a competitive learning network based on a new "conscience" learning algorithm is presented. A number of algorithms for competitive learning networks are compared to the proposed algorithm. The proposed algorithm is tested with different data sets and is shown to be efficient in obtaining near-optimal results. Clustering results produced by the network are checked by an internal index for cluster validity. We conclude this paper with an application of the proposed network in acoustic imaging segmentation for material characterization.
- Artificial neural network
- Ultrasound material characterization
- Unsupervised learning