1 | | ******************************************************* |
---|
2 | | *************************************************** |
---|
3 | | ********************************************** |
---|
4 | | ****************************************** |
---|
5 | | T H E M O N K' S P R O B L E M S |
---|
6 | | ****************************************** |
---|
7 | | ********************************************** |
---|
8 | | *************************************************** |
---|
9 | | ******************************************************* |
---|
10 | | |
---|
11 | | |
---|
12 | | |
---|
13 | | The MONK's problems are a collection of three binary classification |
---|
14 | | problems over a six-attribute discrete domain. Each training/test data |
---|
15 | | is of the form |
---|
16 | | |
---|
17 | | <name>: <value1> <value2> <value3> <value4> <value5> <value6> -> <class> |
---|
18 | | |
---|
19 | | where <name> is an ASCII-string, <value n> represents the value of |
---|
20 | | attribute # n, and <class> is either 0 or 1, depending on the class |
---|
21 | | this example belongs to. The attributes may take the following values: |
---|
22 | | |
---|
23 | | attribute#1 : {1, 2, 3} |
---|
24 | | attribute#2 : {1, 2, 3} |
---|
25 | | attribute#3 : {1, 2} |
---|
26 | | attribute#4 : {1, 2, 3} |
---|
27 | | attribute#5 : {1, 2, 3, 4} |
---|
28 | | attribute#6 : {1, 2} |
---|
29 | | |
---|
30 | | Thus, the six attributes span a space of 432=3x3x2x3x4x2 examples. |
---|
31 | | |
---|
32 | | |
---|
33 | | |
---|
34 | | /*********************************************************************\ |
---|
35 | | *********************************************************************** |
---|
36 | | \*********************************************************************/ |
---|
37 | | |
---|
38 | | |
---|
39 | | |
---|
40 | | The "true" concepts underlying each MONK's problem are given by: |
---|
41 | | |
---|
42 | | MONK-1: (attribute_1 = attribute_2) or (attribute_5 = 1) |
---|
43 | | |
---|
44 | | MONK-2: (attribute_n = 1) for EXACTLY TWO choices of n (in {1,2,...,6}) |
---|
45 | | |
---|
46 | | MONK-3: (attribute_5 = 3 and attribute_4 = 1) or |
---|
47 | | (attribute_5 != 4 and attribute_2 != 3) |
---|
48 | | |
---|
49 | | (with "!=" denoting inequality). |
---|
50 | | MONK-3 has 5% additional noise (misclassifications) in the training set. |
---|
51 | | |
---|
52 | | |
---|
53 | | |
---|
54 | | /*********************************************************************\ |
---|
55 | | *********************************************************************** |
---|
56 | | \*********************************************************************/ |
---|
57 | |
---|
58 | 0, 1 |classes |
---|
59 | |
---|
60 | attribute#1 : 1, 2, 3 |
---|
61 | attribute#2 : 1, 2, 3 |
---|
62 | attribute#3 : 1, 2 |
---|
63 | attribute#4 : 1, 2, 3 |
---|
64 | attribute#5 : 1, 2, 3, 4 |
---|
65 | attribute#6 : 1, 2 |
---|