Considerations of dimensionality versus sample size

It has been shown that the error rate on the design data is a monotonically increasing function of the ratio of sample size to feature size (N/L). Ratios of N/L > 3 are generally tolerable and ratios of N/L > 10 are generally adequate for design stability. It is felt that the results presented here should be kept in mind by those researchers who wish to demonstrate the potential of their pattern recognition algorithm using performance on a sample design set as the demonstration vehicle.