Introduction to Pattern Recognition: A Matlab Approach

160 CHAPTER 7 Clustering

Both are “sensible,” in the sense that the vectors close to each other are included in the same cluster.

However, there is no extra data information to indicate which should be finally selected. In general, the

best approach when dealing with clustering problems is to consult an expert in the field of application.

Another possibility would be the clustering 

3

={{x

1

, x

7

},{x

3

, x

4

},{x

5

, x

6

, x

2

}}, but this is not

sensible since grouping x

1

and x

7

, which are far from each other, seems to contradict physical reasoning.

The same holds true for clustering x

2

together with x

5

, x

6

.

Remarks

• The previous definition of clustering is a loose ...