Fig. 3From: Deep neural networks for wearable sensor-based activity recognition in Parkinson’s disease: investigating generalizability and model complexitytSNE plots with DANN applied for the PAMAP2-to-MHEALTH test case using Chen/Jordao architecture [10, 16]. A Augmented PAMAP2 as the source data, and B reoriented MHEALTH as the target data. C Confusion matrix on target MHEALTH, where the labels represent classes: standing, sitting, lying, walking, climbing stairs, cycling, and runningBack to article page