70 4. FRONTIERS OF GVGAI PLANNING
3 phases 30%
70%
Feature correlation
ϕ
1
ϕ
1
ϕ
2
ϕ
2
ϕ
3
ϕ
4
ϕ
5
ϕ
6
ϕ
7
ϕ
8
ϕ
9
ϕ
10
ϕ
11
ϕ
12
ϕ
1
1.00
0.75
0.50
0.25
0.00
-0.25
-0.50
-0.75
-1.00
1.00
0.75
0.50
0.25
0.00
-0.25
-0.50
-0.75
-1.00
ϕ
2
ϕ
3
ϕ
4
ϕ
5
ϕ
6
ϕ
7
ϕ
8
ϕ
9
ϕ
10
ϕ
11
ϕ
12
ϕ
3
ϕ
4
ϕ
5
ϕ
6
ϕ
7
ϕ
8
ϕ
9
ϕ
10
ϕ
11
ϕ
12
ϕ
1
ϕ
2
ϕ
3
ϕ
4
ϕ
5
ϕ
6
ϕ
7
ϕ
8
ϕ
9
ϕ
10
ϕ
11
ϕ
12
2
6
4.4. GENERAL WIN PREDICTION IN GVGAI 71
4.4.2 PREDICTIVE MODELS
80
20
precision D
TP
TP C FP
;
recall D
TP
TP CFN
;
F1 D 2
precision recall
precision C recall
;
TP FP
FN
25%
Baseline Model
Oy D
8
<
:
win if
3
>
4
lose otherwise:
0:59 0:70
R
g
Precision Recall F1-Score Support
Loss 0.83 0.52 0.64 20,500
Win 0.35 0.70 0.46 7,500
Average/Total 0.70 0.57 0.59 28,000
72 4. FRONTIERS OF GVGAI PLANNING
Classifier Selection—Global Model
5 5
5 10 1
10
0:95 1:00 0:98 0:96 1:00 0:95 0:66
Precision Recall F1-Score Support
Loss 1.00 0.99 0.99 20,500
Win 0.97 0.99 0.98 7,500
Average/Total 0.99 0.99 0.99 28,000
φ
1
0.24
φ
2
0.04
φ
3
0.08
φ
4
0.06
φ
5
0.2
φ
6
0.1
φ
7
0.12
φ
8
0
φ
9
0.06
φ
10
0.02
φ
11
0.02
φ
12
0.06
Model Training
30%
30%
E
g
M
g
L
g
4.4. GENERAL WIN PREDICTION IN GVGAI 73
20
10 0:80 0:82
0:99
0:73 0:80 0:99 0:70 0:80 0:99
Live Play Results
live
T D f100 a W 8 a 2 Œ1; 20 W a 2 Ng
t 2 T 14
20 20 5
100
Aliens Defem Chopper Eggomania
R
g
Ghost Buster Color Escape Frogs
E
g
M
g
L
g
Fireman
E
g
M
g
0%
Defem Ghost Buster
50% 0:8
M
g
do 50% Defem
Ghost Buster
74 4. FRONTIERS OF GVGAI PLANNING
Win Average = 0.99 (+/- 0.08) Win Average = 0.13 (+/- 0.33) Win Average = 0.88 (+/- 0.32)
Win Average = 0.00 (+/- 0.00) Win Average = 0.03 (+/- 0.18) Win Average = 0.93 (+/- 0.25)
Win Average = 0.08 (+/- 0.28) Win Average = 0.00 (+/- 0.03) Win Average = 0.41 (+/- 0.49)
Win Average = 0.00 (+/- 0.00)
Model F1-score
Win Average = 0.00 (+/- 0.00) Win Average = 0.00 (+/- 0.00)
1.0
0.8
0.6
0.4
0.2
0.0
Model F1-score
1.0
0.8
0.6
0.4
0.2
0.0
Model F1-score
1.0
0.8
0.6
0.4
0.2
0.0
Model Accuracy
1.0
0.8
0.6
0.4
0.2
0.0
Model F1-score
1.0
0.8
0.6
0.4
0.2
0.0
Model F1-score
1.0
0.8
0.6
0.4
0.2
0.0
Model F1-score
1.0
0.8
0.6
0.4
0.2
0.0
Model F1-score
1.0
0.8
0.6
0.4
0.2
0.0
Model F1-score
1.0
0.8
0.6
0.4
0.2
0.0
Model Accuracy
1.0
0.8
0.6
0.4
0.2
0.0
Model Accuracy
1.0
0.8
0.6
0.4
0.2
0.0
Model Accuracy
1.0
0.8
0.6
0.4
0.2
0.0
200 400 600 800 1000
200 400 600 1200800 1000 1400 1600
200100 300 400
Game Tick
Game Tick
250 500 750 1000 1250 1500 1750 2000
Game Tick
250 500 750 1000 1250 1500 1750 2000
Game Tick
200 400 600 1200800 1000 1400 1600
Game Tick
250 500 750 1000 1250 1500 1750 2000
Game Tick
250 500 750 1000 1250 1500 1750 2000
Game Tick
250 500 750 1000 1250 1500 1750 2000
Game TickGame Tick
200 400 600 800 1000 1200 1400
Game Tick
250 500 750 1000 1250 1500 1750 2000
Game Tick
250 500 750 1000 1250 1500 1750 2000
Game Tick
600500 700 800
14
100 X
0
Aliens Boulderdash Butterflies Caky Baky Chase Chopper Color Escape Cops Defem Deflection
DigDug Donkeykong
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.