Scatter Plot from Scientific Research

To estimate the prediction errorσActual(which we are not allowed to directly compute becauseV0’s titers are withheld), we define the transferability relationf2→1between the training errorσTrainingin dataset 2 and actual errorσActualin dataset 1 using the decision trees that predict virusesV1–Vn(without usingV0). Applying this relation to the training error,f2→1(σTraining), estimatesσActualforV0.

DOI