Skip to main content
. 2022 Jun 15;8(24):eabn7630. doi: 10.1126/sciadv.abn7630

Fig. 3. Semisupervised node classification on three benchmark graph databases.

Fig. 3.

(A) Test accuracy convergence plots of DGNN-O with optical DPU output classifier. (B) Test accuracy convergence plots of DGNN-E with electronic output classifier. (C) t-distributed stochastic neighbor embedding (t-SNE) visualization of node representations of DGNN-E on the Amazon Photo dataset. (D to F) Confusion matrices of DGNN-E classification result on three graphs with binary modulation and system errors by including Gaussian noise with an SD of 0.3.