N

D

C

Q

E

G

H

I

L

K

M

F

P

S

T

W

Y

V

2

2

2

1

2

1

1

3

3

2

2

3

2

1

1

3

2

2

(a) (b)

The comparison between the two models. The amino acids enclosed by three

atterns (the circles, the triangles and the diamonds) represent three clusters. (a)

ns model. (b) The hierarchical cluster model.

rarchical cluster model for a data set is always deterministic

this clustering algorithm does not involve a process of

ing model parameters at the beginning of a learning process.

, a K-means model is not deterministic. Most machine learning

how different modelling outcomes for the same data set using the

del parameters in different runs. This is because they start from a

process with random values assigned to the model parameters.

cess is called the model parameter initialisation. The cluster

f a K-means model are also assigned by random values at the

g of a learning process. Because of this, two data points may be

into the same cluster in one model, but may be classified into

clusters in the other model using the same K-means algorithm.

26 shows four K-means models for a data set with three clusters

oints. They show some different clustering results. Therefore, it

d to take care when clustering data using the K-means algorithm.