6.1. Estimating the Connectivity Matrix
In terms of performance in the presence of additive noise, Figure 4 depicts that NINCA and PosNCA achieve a higher degree of accuracy when the SNR is high. Moreover, the performance of ROBNCA and NCA is also accurate and consistent compared to FastNCA. When the data are corrupted with both noise and outliers, according to Figure 6, ROBNCA achieves the best performance against outliers. The NSME of NINCA and PosNCA increases significantly compared to the case without outliers, especially when the SNR is high. Even though the performance of FastNCA does not degenerate when outliers exist, the NMSE is still relatively large compared to other algorithms.