Gradient Neural Dynamics Based on Modified Error Function
Table 1: Experiment results for squared matrices
| n |
Rank A |
Rank B |
Time |
GNN |
GGNN |
γ |
GNN CPU Time |
GGNN CPU Time |
| 50 |
2 |
2 |
10-6 |
19.38 |
19.38 |
100 |
20.92 |
14.45 |
| 50 |
2 |
50 |
10-6 |
17.29 |
17.13 |
100 |
9.05 |
15.17 |
| 50 |
50 |
50 |
10-6 |
14.75 |
14.23 |
100 |
8.76 |
15.30 |
| 50 |
10 |
10 |
10-6 |
15.77 |
15.59 |
100 |
7.58 |
16.95 |
| 50 |
10 |
50 |
10-6 |
15.44 |
15.11 |
100 |
7.08 |
14.79 |
| 50 |
50 |
50 |
10 |
2.30 |
4.022 |
100 |
29.61 |
28.10 |
| 50 |
2 |
2 |
10 |
NRT |
19.30 |
100 |
NRT |
20.38 |
| 10 |
10 |
10 |
10-2 |
2.77 |
2.59 |
1 |
8.82 |
1.43 |
| 10 |
2 |
10 |
10-2 |
3.36 |
3.34 |
1 |
16.22 |
1.43 |
| 10 |
2 |
2 |
10-2 |
3.60 |
3.54 |
1 |
9.01 |
1.38 |
| 10 |
5 |
10 |
10-2 |
3.09 |
2.96 |
1 |
9.78 |
1.34 |
| 10 |
5 |
5 |
10-2 |
3.22 |
3.13 |
1 |
9.92 |
1.47 |
< Back to article
