Distribution of the validation mean absolute errors of the models trained during the hyperparameter tuning. Left panel: dependence on the regularisation factor (pink), with f = 10⁻⁶ indicated as the optimal value. Middle panel: dependence on the learning rate (pink). The distribution highlights the trials with the optimal f value (green) and indicates optimal performance for r = 0.01. Right panel: dependence on the number of units (pink). The distribution highlights the trials with the optimal f (green) and r values (blue). The green violin plots focus on the models with the best performing regularisation factor, f = 10⁶, while the blue ones also focus on those with the best performing learning rate, r = 0.01.