Fig. 1: Machine learning model performance.
From: Sparse machine learning of resting-state fMRI reveals brain-wide dysconnectivity in hyperacusis
a Hyperacusis classification performance on test data of n = 209 leave-one-out model repetitions. Points represent scores for each model, bars represent the mean values. F1-score = 0.679 ( ± 0.021), precision = 0.572 ( ± 0.022), and recall = 0.835 ( ± 0.024). Error bars represent standard errors (standard deviation/no. of model repetitions) (b) Mean performance measures for 500 null models (histograms) compared against mean performance of true model (dashed lines) for F1-score (left panel; gray), precision (middle panel; blue), and recall (right panel; red). See “Methods—Machine learning classification—Model training and accuracy evaluation” for details.
