Table 2.

Results of the low-level HAR model on the hold-out set. Mean and std across the 10 participants in the holdout set for the different performance metrics.

	LogLoss	Balanced accuracy	F1-macro	F1-micro
	$0.62\pm 0.17$	$0.67\pm 0.12$	$0.52\pm 0.08$	$0.74\pm 0.08$