Table 1 The performance of reconstructing different types of artificial networks.
From: Reconstructing direct and indirect interactions in networked public goods game
N | 〈k〉 | σ | N h | ER | WS | NW | BA |
|---|---|---|---|---|---|---|---|
| Â | 4 | 0 | 0 | 0.61 | 0.41 | 0.50 | 0.95 |
| Â | 4 | 0.05 | 0 | 0.64 | 0.48 | 0.49 | 1.20 |
100 | 4 | 0.5 | 0 | 1.35 | 1.80 | 1.74 | 5.02 |
| Â | 6 | 0 | 0 | 0.89 | 0.75 | 0.77 | 1.10 |
| Â | 8 | 0 | 0 | 1.06 | 0.97 | 1.03 | 1.15 |
300 | 4 | 0 | 0 | 0.24 | 0.2 | 0.25 | 0.71 |
500 | 4 | 0 | 0 | 0.20 | 0.15 | 0.19 | 0.46 |
100 | 4 | 0 | 1 | 1.81 | 0.77 | 0.51 | 1.51 |
- Data amount needs to achieve 90% success rates (SR) for four artificial network models, where SR = TPR × TNR (SR is area under ROC with given threshold, see Supplementary Figure S1 for more details). ER, WS, NW and BA networks with different network size N, average degree 〈k〉 and measurement noise (Gaussian white noise
) are considered. Nh denotes the number of the hidden nodes. The results are obtained by averaging over 10 independent realizations. More details of the success rates as a function of data amount for different cases can be found in Supplementary Figures S2–S5.
) are considered. Nh denotes the number of the hidden nodes. The results are obtained by averaging over 10 independent realizations. More details of the success rates as a function of data amount for different cases can be found in 