Title :
Learning with Lotto-type competition
Author :
Luk, Andrew ; Lien, Sandra
Author_Institution :
St. B&P Neural Investments Pty. Ltd., Wesleigh, NSW, Australia
Abstract :
Classical winner-takes-all competitive learning can be modified such that not only the winner is rewarded but also all the losers are penalised. The idea is similar to a simplified Lotto system where all losers are divested of their bets. Experimental results indicate that convergence occurs only if parameters are suitably set
Keywords :
convergence; neural nets; unsupervised learning; Lotto-type competition; competitive learning; convergence; neural networks; winner-takes-all learning; Algorithm design and analysis; Australia; Cameras; Clustering algorithms; Convergence; Investments; TV; Time series analysis;
Conference_Titel :
Neural Networks Proceedings, 1998. IEEE World Congress on Computational Intelligence. The 1998 IEEE International Joint Conference on
Conference_Location :
Anchorage, AK
Print_ISBN :
0-7803-4859-1
DOI :
10.1109/IJCNN.1998.685933
