Table 3.

Performance of a linear classifier identifying amplifier constructors from maximal remeeting time $max({\tau }_i)$ and spectral gap ${\lambda }_2$ for cubic and quartic regular graphs

N	k
	3		4
	PRE	ACC	PRE	ACC
12	100	96.4706	100	99.8705
13			95.8333	99.9351
14	41.1765	95.6778	90.8333	99.9297
15			85.5403	99.3778
16	23.2044	96.4286	70.5524	99.8950
18	20.6547	96.8887
20	14.5562	95.2442
22	6.1967	97.7082

The results are the precision PRE and the accuracy ACC in $\%$. The precision PRE is defined as the fraction of all correctly classified amplifier constructors (true positives over all graphs identified as amplifier constructor), while the accuracy is the fraction of all correctly classified graphs (true positives plus true negatives over all graphs)