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

<div>Test sample points of two supervised methods and HTF-PSA-SSL after dimensionality reduction using t-SNE under 0 dB signals. (a) Supervised (100%) and (b) HTF-PSA-SSL are well-aggregated. (c) Supervised (1%) is heavily scattered.</div>

Wireless Communications and Mobile Computing

fig11

Figure 11

Figure 11: Semi-supervised Learning for Automatic Modulation Recognition Using Haar Time–Frequency Mask and Positional–Spatial Attention 