Figure 3.

Multilayer perceptron (MLP). Our MLP with seven neurons in input layer, consequently features chosen by dimensionality reduction methods, six neurons in hidden layer, chosen by grid search method, classify our dataset in output layer to predict pathological or healthy subjects’ gait: CV, coefficient of variation of step length; HR ap, harmonic ratio anteroposterior; RQAdet_ap, % determinism in the recurrence quantification analysis anteroposterior; pwPD, people with Parkinson’s disease; HS, healthy subjects.