Semi-supervised Learning for Automatic Modulation Recognition Using Haar Time–Frequency Mask and Positional–Spatial Attention

Table 2

Computation complexity.

Network	FLOPs	Parameters	Memory
SVM [47]	—	54.2 K	1.5 M
RF [48]	—	63.7 K	3.5 M
SSRCNN [19]	390 K	52.8 K	227.7 K
EDCT [31]	9 M	291 K	1.2 M
SimAMC [33]	25 M	620 K	2.5 M
CNN5	2.5 G	1.66 M	6.7 M
CNN5 + spatial	2.5 G	1.66 M	6.7 M
SE [40]	2.5 G	1.77 M	7.1 M
Fca [41]	2.5 G	1.85 M	11.2 M
CBAM [42]	2.5 G	1.77 M	7.1 M
CNN5 + positional	2.53 G	1.82 M	7.4 M
HTF-PSA-SSL	2.53 G	1.82 M	7.4 M

Note. Bold values indicate the highest value of FLOPs, Paramters and Memory.