Samplings per Experiment | Algorithms |
ε
M
|
ε
S
|
ε
F
|
---|
| | Mean | STD | Mean | STD | Mean | STD |
4 | LS | 1.34 | 0.34 | 1.01 | 0.14 | 0.04 | 0.01 |
| TLS | 1.34 | 0.34 | 1.01 | 0.14 | 0.04 | 0.01 |
| CTLS | 1.34 | 0.34 | 1.01 | 0.14 | 0.04 | 0.01 |
8 | LS | 0.95 | 0.27 | 0.50 | 0.02 | 0.03 | 0.01 |
| TLS | 25.03 | 196.73 | 1.01 | 0.15 | 1.06 | 8.42 |
| CTLS | 0.44 | 0.12 | 0.50 | 0.00 | 0.02 | 0.00 |
12 | LS | 0.61 | 0.08 | 0.50 | 0.00 | 0.02 | 0.00 |
| TLS | 47.53 | 241.86 | 0.92 | 0.31 | 2.49 | 13.07 |
| CTLS | 0.49 | 0.06 | 0.50 | 0.02 | 0.02 | 0.00 |
16 | LS | 0.44 | 0.06 | 0.50 | 0.00 | 0.02 | 0.00 |
| TLS | 50.96 | 833.28 | 1.02 | 0.20 | 3.11 | 50.06 |
| CTLS | 0.49 | 0.05 | 0.50 | 0.02 | 0.02 | 0.00 |
- The table shows the error comparisons in terms of the mean and the standard deviation (STD) for different number of data for each method based on 1000 Monte-Carlo Simulations. The measurements are taken every 2 hours and the states converge to steady states around the 16-th sample. ε
M
is the sum of two tems, i.e (1/N1) Σ |α
i j
| and (1/N2) Σ |β
i j
|, where α
i j
and β
i j
are the relative magnitude errors in the non-zero and zero elements of the true Jacobian, respectively, and N1 and N2 are the number of non-zero and zero elements in the true Jacobian, respectively. ε
S
is given by (1/n2) Σ |sign (
i j
) - sign (f
i j
)|, i.e. the average sign differences, where
i j
and f
i j
are the (i-th row, j-th column) elements of the estimated and the true Jacobian, respectively. ε
F
is the Frobenius norm of the difference between the estimated and the true Jacobian, i.e. || - F||
F
.