Table 5.

Table showing a performance comparison between the top performing networks from each network type, evaluated on the test set, along with the best performing LSTM network from Pettersson and Falkman (2022).

Network	Th L	AP (%)	AM (%)	AVP (%)	P	T[ms]
Best networks from each architecture—Test set
Enc1	0.53	82.74	80.06	89.10	7.02k	31.43 ± 4.05
Enc2	0.63	79.12	65.77	87.73	5.79k	30.26 ± 3.89
Enc3	0.48	69.78	71.29	82.56	5.25k	29.96 ± 3.91
Mix1	0.43	79.09	70.51	86.90	6.77k	29.30 ± 3.89
Mix2	0.42	80.19	72.28	87.53	6.57k	29.37 ± 3.88
Mix3	0.25	76.97	77.86	87.61	4.99k	29.42 ± 3.88
LSTM	0.38	70.70	67.89	81.29	6.99k	30.39 ± 4.10

The bold values indicate the best (largest) and most interesting value for three accuracy metrics.