Parameterizing neural network models to improve land classification performance - Urban Remote Sensing: Monitoring, Synthesis and Modeling in the Urban Environment - page 94

TABLE 7.4 List of the 53 neural network models configured with various internal parameter settings and classification accuracies obtained.

No. HL AF TT LR MO IT Classification Accuracy (%)

1 0 log-sig 0.2 0.01 0.6 1000 79.33

2 1 log-sig 0.2 0.01 0.6 1000 80.00

3 2 log-sig 0.2 0.01 0.6 1000 77.67

4 3 log-sig 0.2 0.01 0.6 1000 66.67

5 4 log-sig 0.2 0.01 0.6 1000 N/A

6 1 log-sig 0 0.01 0.6 1000 82.00

7 1 log-sig 0.1 0.01 0.6 1000 79.67

8 1 log-sig 0.2 0.01 0.6 1000 80.00

9 1 log-sig 0.3 0.01 0.6 1000 80.33

10 1 log-sig 0.4 0.01 0.6 1000 80.00

11 1 log-sig 0.5 0.01 0.6 1000 79.67

12 1 log-sig 0.6 0.01 0.6 1000 79.00

13 1 log-sig 0.7 0.01 0.6 1000 78.33

14 1 log-sig 0.8 0.01 0.6 1000 79.00

15 1 log-sig 0.9 0.01 0.6 1000 72.00

16 1 tan-sig 0 0.01 0.6 1000 34.67

17 1 tan-sig 0.1 0.01 0.6 1000 34.00

18 1 tan-sig 0.2 0.01 0.6 1000 26.67

19 1 tan-sig 0.3 0.01 0.6 1000 44.67

20 1 tan-sig 0.4 0.01 0.6 1000 44.67

21 1 tan-sig 0.5 0.01 0.6 1000 44.00

22 1 tan-sig 0.6 0.01 0.6 1000 40.67

23 1 tan-sig 0.7 0.01 0.6 1000 38.67

24 1 tan-sig 0.8 0.01 0.6 1000 40.33

25 1 tan-sig 0.9 0.01 0.6 1000 35.00

26 1 log-sig 0 0.3 0.6 1000 53.67

27 1 log-sig 0 0.25 0.6 1000 60.00

28 1 log-sig 0 0.2 0.6 1000 75.00

27 1 log-sig 0 0.15 0.6 1000 64.67

30 1 log-sig 0 0.10 0.6 1000 72.33

31 1 log-sig 0 0.05 0.6 1000 75.33

32 1 log-sig 0 0.01 0.6 1000 82.00

33 1 log-sig 0 0.005 0.6 1000 81.33

34 1 log-sig 0 0.001 0.6 1000 75.00

35 1 log-sig 0 0.01 0 1000 81.33

36 1 log-sig 0 0.01 0.1 1000 81.33

37 1 log-sig 0 0.01 0.2 1000 81.33

38 1 log-sig 0 0.01 0.3 1000 82.67

39 1 log-sig 0 0.01 0.4 1000 82.00

40 1 log-sig 0 0.01 0.5 1000 81.33

41 1 log-sig 0 0.01 0.6 1000 82.00