Figure 3From: Scientific knowledge is possible with small-sample classificationSample-conditioned RMS probability densities. Probability densities for the sample-conditioned RMS of leave-one-out (dashed line) and the Bayesian error estimator (solid line) in a discrete model with b=16 bins, prior probability c=0.5, n=30 training points, and an average true error of 0.25.Back to article page