1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
|
***read README.txt before this file***
this ideas are related to shostakovich prelude n.2 in a minor
how do we map notes <--> genome ?
--------------- 1st idea: whole MIDI, absolute values
keep all the chromatic values, all midi notes
the fitness functions and crossover functions will manage the thing
(no notes outside current chord, etc..)
each gene is a byte,
- first 7 bits for midi note (unsigned char, 0-127)
- last bit for rest (1=play, 0=rest)
why yes:
-not restricted to prelude n.2 musical space
-not restricted to tonal music space
why not:
-huge transition table, intervals table
-we don't want values we don't need
-difficult to write fitness functions that check for tonality, chords, etc..
---------------- 2nd idea: chromatic scale, chord relative
notes are relative to current chord but still in a chromatic space.
the critics must choose the diatonic ones and discard atonal ones
each gene is a byte,
- first 7 bits for midi note (unsigned char, 0-127)
- last bit for rest (1=play, 0=rest)
why yes:
-we can use the same melody to different chords
-not restricted to prelude n.2 musical space
-not restricted to tonal music space
why not:
-still huge tables
-difficult to write fitness functions that check for tonality, chords, etc..
---------------- 3rd idea: diatonic scale, chord relative
the musical space is diatonic, no more chromatic. critics need to work much less,
they only have to know if a note is in the chord or not.
problem: how to use chromatic passing notes this way?
possibile notes are 7*4=28 (7 notes per octave) + passing notes
we could use : 1 byte for each gene
- 5 bits for note (0-32)
- 2 bit for cromatic passing note (0=no, 1=descending, 2=ascending, 4=not used)
- 1 bit for rests
why yes:
-we can use the same melody to different chords
-quick fitness functions
-we don't have notes we don't want
why not:
-restricted to tonal music space
---------------- 4th idea: chord notes
the musical space are the chord notes, not even the diatonic scale.
critics don't have to check anything!
problem: how to use (chromatic or diatonic) passing notes?
possibile notes are 3*4=12 (3 notes per octave) + passing notes
we should use a struct for the gene:
{
- note (from 0 to 16)
- rest (0/1)
- passing_note (from -4 to +4)
}
possible passing notes are 9:
no passing, from -4 semitones to +4 semitones
the fitness functions and crossover functions will have to care sbout rules for passing notes:
+3 and +4 are allowed only for the 3rd note of the chord
-4 and -3 are allowed only for the 1st note of the chord
every passing note can exist only if the note before was next to it and proceeding in the same direction
each passing note must resolve on the nearer chord note in the right direction
why yes:
-we can use the same melody to different chords
-quickest fitness functions
-we don't have notes we don't want
why not:
-restricted to tonal music space
-restricted to 3-notes-chords-based harmonies (jazz is 4 notes based)
common problems:
i can think a way to rule passing notes (a passing note must come from a consonant note and go to a consonant note)
but what can we do if a passing not is at the very end of the melody? (i don't know what will the next note be)
...we should have to know what will be the next chord...
Davide Morelli
17.09.05
|
