Title :
Generating data sets for benchmarking
Author_Institution :
Dept. of Comput. Sci., Tasmania Univ., Tas., Australia
Abstract :
A new method of benchmarking neural networks based on Voronoi diagrams is introduced. Their complexity is examined and it is shown that data sets of increasing difficulty may be generated. Experiments are conducted examining the performance of five known classification methods on examples of Voronoi data sets
Keywords :
computational geometry; learning by example; neural nets; pattern classification; Voronoi diagrams; benchmarking; classification methods; complexity; neural networks; Artificial neural networks; Benchmark testing; Character generation; Computer science; Euclidean distance; Learning systems; Neural networks; Power generation; Random number generation; System testing;
Conference_Titel :
Neural Networks, 1995. Proceedings., IEEE International Conference on
Conference_Location :
Perth, WA
Print_ISBN :
0-7803-2768-3
DOI :
10.1109/ICNN.1995.489010
