1. Adaptive Games
Content Generation
“Mario”
Mohammad Shaker
Department of Artificial Intelligence
IT University of Damascus
Seminar of Artificial Neural Networks
ZGTR

2. Outline
• Readings
• Motivation
• The proposed approach
• Experiments
• ANN Implementation
• Results
• Conclusion

3. Readings
• Towards Automatic Personalized Content
Generation for Platform Games
Noor Shaker, Georgios N. Yannakakis, Member, IEEE, and Julian Togelius,
Member, IEEE
• Feature Analysis for Modeling Game Content
Quality
Noor Shaker, Georgios N. Yannakakis, Member, IEEE, and Julian Togelius,
Member, IEEE

4. Motivation

5. Motivation

6. Motivation

7. Motivation

8. Motivation

9. Motivation

10. Motivation

11. The Big Picture

12. The Big Picture
Game Player

13. The Big Picture
Game Player

14. The Big Picture
Game Player
Player Experience
Model

15. The Big Picture
Game Player
Player Experience
Model
Game Adaptation

16. The Big Picture
Game Player
Player Experience
Model
Game Adaptation

17. The Big Picture
Game Player
Player Experience
Model
Game Adaptation

18. The Big Picture
Game Player
Player Experience
Model
Game Adaptation

19. The Game

20. The Game

21. Open Questions!
 Session period? (frequency of adaptation)
 The most useful information about game content?
 Game aspects with major affect on player
experience?

22. Open Questions!
 Session period? (frequency of adaptation)
 The most useful information about game content?
 Game aspects with major affect on player
experience?

23. Approach
Design

24. Approach
Design Collect
Data

25. Approach
Model
Design Collect Player’s
Data Emotion

26. Data Collection
 40 small levels
(one-third of usual size)
 600 game pairs
 Features
 Six controllable features
 Players preferences of engagement

27. Data Collection
 40 small levels
(one-third of usual size)
 600 game pairs
 Features
 Six controllable features
 Players preferences of engagement

28. Data Collection - Controllable Features
 number of gaps
 average width of gaps
 number of enemies
 number of powerups
 number of boxes
 Enemies placement
 Around horizontal boxes
 Around gaps
 Random placement

29. Experiments

30. Experiment 1
 How long the game session should be in order to be
able to extract useful information?

31. Experiment 1
 How long the game session should be in order to be
able to extract useful information?

32. Segmentation

33. Segmentation

34. Content-Driven Preference Learning
• It’s the use of genetic algorithms to evolve the
weight of neural networks to learn preference data.

35. Content-Driven Preference Learning
• It’s the use of genetic algorithms to evolve the
weight of neural networks to learn preference data.
Levels

36. Content-Driven Preference Learning
• It’s the use of genetic algorithms to evolve the
weight of neural networks to learn preference data.
Levels Segmentation

37. Content-Driven Preference Learning
• It’s the use of genetic algorithms to evolve the
weight of neural networks to learn preference data.
Feature
Levels Segmentation
extraction

38. Content-Driven Preference Learning
• It’s the use of genetic algorithms to evolve the
weight of neural networks to learn preference data.
NeuroEvolutionary
Feature preference
Levels Segmentation
extraction learning

39. Content-Driven Preference Learning
• It’s the use of genetic algorithms to evolve the
weight of neural networks to learn preference data.
NeuroEvolutionary
Feature Player’s
Levels Segmentation preference
extraction Engagement
learning

40. Content-Driven Preference Learning
Feature
extraction
NeuroEvolutionary
preference
learning

41. Feature Feature Feature
extraction extraction extraction
NeuroEvolutionary NeuroEvolutionary NeuroEvolutionary
preference preference preference
learning learning learning

42. Experiment 2
 How can we extract the most useful information
about game content?

43. Experiment 2
 How can we extract the most useful information
about game content?

44. Game Content Representation
 Statistical features
 Sequences

45. Game Content Representation
 Statistical features
 Six controllable features
 Used for level generation
 Sequences

46. Game Content Representation
 Statistical features
 Six controllable features
 Used for level generation
 Sequences
 Numbers representing different types of game content
o Platform structure, S
o Enemies placement, Ep
o Enemies and items placement, D

47. Sequence Mining

48. Sequence Mining

49. Sequence Mining-SPADE
SPADE occurrences
Frequent
40 levels Subseq.
seq.

50. Content-Driven Preference Learning
ANN-
Statistical NeuroEvolutionary Player’s
features Preference Engagement
Learning
ANN-
Sequential NeuroEvolutionary Player’s
features Preference Engagement
Learning

51. Experiment 3
 What are the game aspects that have the major
affect on player experience?

52. Experiment 3
 What are the game aspects that have the major
affect on player experience?

53. Content-Driven Preference Learning
Statistical ANN-
features
Feature NeuroEvolutionary Player’s
Sequential selection Preference Engagement
features Learning

54. ANN Implementation

55. ANN Implementation
• Multilayer perceptrons (MLPs)
o ANN inputs
• Controllable features
• Sequences as features
o ANN output
• Value of the engagement preference

56. ANN Training
• Genetic algorithms (GAs)
o No prescribed target outputs

57. ANN Training
• Genetic algorithms (GAs)
o No prescribed target outputs
• How it works?

58. ANN Training
• Genetic algorithms (GAs)
o No prescribed target outputs
• How it works?
players’ magnitude of
reported corresponding
emotional model (ANN)
preferences output

59. ANN Training
• Genetic algorithms (GAs)
o No prescribed target outputs
• How it works?
players’
reported
emotional
preferences - magnitude of
corresponding
model (ANN)
output

60. ANN Implementation

61. ANN Implementation
SF
CF

62. ANN Implementation
SF
CF

63. Optimizing Neural Networks Topologies
• 2 hidden layers (Max.)

64. Optimizing Neural Networks Topologies
• 2 hidden layers (Max.)
• Multiple experiments
 1 hidden layer, Adding two neurons at each step
 2 neurons - 8 neurons

65. Optimizing Neural Networks Topologies
• 2 hidden layers (Max.)
• Multiple experiments
 1 hidden layer, Adding two neurons at each step
 2 neurons - 8 neurons
 2 hidden layers, Adding two neurons at each step
 1st Hidden layer
 2 neurons - 10 neurons
 2nd Hidden layer
 2 neurons - 8 neurons

66. Optimizing Neural Networks Topologies
• 2 hidden layers (Max.)
• Multiple experiments
 1 hidden layer, Adding two neurons at each step
 2 neurons - 8 neurons
 2 hidden layers, Adding two neurons at each step
 1st Hidden layer
 2 neurons - 10 neurons
 2nd Hidden layer
 2 neurons - 8 neurons

67. ANN Adaptation

68. ANN Implementation
SF
CF

69. ANN Adaptation
SF Prediction of
player’s
CF emotion

70. ANN Adaptation
SF Prediction of
player’s
CF emotion
Gaps #: 4-10
Gaps width: 10-30
Gaps placement: 0-1
Switch:0-1

71. ANN Adaptation
SF Prediction of
player’s
CF emotion
Exhaustive
search
Gaps #: 4-10
Gaps width: 10-30
Gaps placement: 0-1
Switch:0-1

72. ANN Adaptation
SF Prediction of
player’s
CF emotion
Exhaustive
search
Gaps #: 4-10
Gaps width: 10-30
Gaps placement: 0-1
Switch:0-1

73. ANN Adaptation
level1 level2 level20 level21 level50
Adapt Adapt Adapt Adapt

74. Neural Networks Input Representation
Statistical ANN-
features
Feature NeuroEvolutionary Player’s
Sequential selection Preference Engagement
features Learning

75. Game Content Representation
Statistical
features
Feature
Sequential selection
features

76. Game Content Representation

77. Game Content Representation
The best-performing MLP models evaluated on occurrences
of frequent subsequences of length three extracted from the 40 levels

78. MLPs Performance on Full Information
about Game Content
The topology and performance of the best MLP models evaluated on full and
partial information about game content. the MLP performance presented is the
average performance over 20 runs.

79. Results
The performance and topologies of MLP models evaluated on full and partial
information of game content using statistics from the game window and from two
and three segments to which the window has been divided. The performance
presented is the average over five runs.

80. Content-Driven Preference Learning
Statistical ANN-
features
Feature NeuroEvolutionary Player’s
Sequential selection Preference Engagement
features Learning

81. Conclusion
 Combining both sequential and statistical features
gives better results in predicting players' reported
emotional state.
 Partitioning the level causes a significant decrease
(p < 0.05) in the accuracy of predicting player’s
reported engagement. This suggests that there
might be information loss because of decomposing
the data and that this loss causes a performance
decrease.
 Multiple perspectives can be done in reference to
this study which is already going on!

82. Thank you!