The authors present a new, fast computation algorithm which significantly reduces the training time of the competitive learning (CL) algorithm without sacrificing its performance. The training process is performed in the wavelet domain to identify the winning neural units with a small number of multiplications. Simulation results show that in many cases the algorithm requires < 10% of the computational time required by traditional CL algorithms.
- Learning systems
- Neural networks
- Wavelet transforms
ASJC Scopus subject areas
- Electrical and Electronic Engineering