Results from “Model misspecification” experiments (see Table 1).

The name of each experiment follows this convention: experiment_exp_{rate}_{number of taxa}_{substitution model of testing dataset/substitution model of training dataset}. The “rate” corresponds to the rate parameter of a uniform distribution. The following values were computed for each branch and tree length: mean squared error (MSE); mean absolute error (MAE); Spearman’s correlation coefficient (rho); Spearman’s correlation test P value (P_rho); bias (Bias); Kolmogorov-Smirnov test statistic D (D); Kolmogorov-Smirnov test P value (P_D). CNN = convolutional neural network; CNN- ROE = convolutional neural network–regression of observed on estimated values; MLP = multilayer perceptron; MLP-ROE = multilayer perceptron–regression of observed on estimated values; ML = maximum likelihood. (XLSX)