Calculate the quantization error for the network below given instances, 2 and 3.

                1             2             3             4            

1             1.2         1.4         2.4         3.0

2             1.4         2.2         2.6         3.2

3             1.8         2.7         2.6         3.4

4             2.0         2.8         2.8         3.5

Consider the self-organizing map belo, which shows the h values calculated using a Gaussian neighbourhood function. Are the h values correctly calculated? Assume the learning rate is one and sigma is one.

                1             2             3

1             0.268    0.607    0.268

2             0.607    1.000    0.607

3             0.268    0.607    0.268

Consider a data set that consists of eight instances. The table below shows the Euclidean distance between instances. Which instance should be selected as w11 when using hypercube initialisation?

      1      2       3       4       5       6       7       8      

1    0      

2   1,25   0     

3    2        1,25   0

4    1,16     0,01   1,16   0

5    1,303   0,553   0,203  0,463    0

6    1         1,25      3        1,36       2,3025      0

7    0,423  1,023    2,423   1,063     1,725       0,123     0

8    0,75      0,5      0,75      0,51      0,503      0,75      0,523      0

True/False. When a conscience factor is used, objects might be “further” than they appear?

True/False. LVQ1 training is based on competitive learning?