Motivate AI Class with Interactive Computer Game

18
Motivate AI Class with Interactive Computer Game Author Akcell Chiang Presented by Yi Cheng Lin

description

Motivate AI Class with Interactive Computer Game. Author : Akcell Chiang Presented by Yi Cheng Lin. Outline. Introduction Case-based Reasoning (CBR) System Architecture Experiment Evaluation Conclusion. Introduction. - PowerPoint PPT Presentation

Transcript of Motivate AI Class with Interactive Computer Game

Page 1: Motivate AI Class with Interactive Computer Game

Motivate AI Class with Interactive Computer Game

Author : Akcell ChiangPresented by Yi Cheng Lin

Page 2: Motivate AI Class with Interactive Computer Game

Outline

Introduction Case-based Reasoning (CBR) System Architecture Experiment Evaluation Conclusion

Page 3: Motivate AI Class with Interactive Computer Game

Introduction Commercial interactive games is one of the

major entertainment in our society, the game annual expense is more than the film industry for a family in average

This paper reports adapting traditional Pacman game with Machine Learning technology Case-based Reasoning (CBR) to provide student learning motivation in the AI subject teaching

Page 4: Motivate AI Class with Interactive Computer Game

Case-based Reasoning (CBR)

Case-based reasoning (CBR) is the process of solving new problems based on the solutions of similar past problems

Page 5: Motivate AI Class with Interactive Computer Game

System Architecture

Page 6: Motivate AI Class with Interactive Computer Game

environment Brick puzzle world, the whole system running-map consists of a 15x15 rectangular grid,

Page 7: Motivate AI Class with Interactive Computer Game

System architecture diagram

Page 8: Motivate AI Class with Interactive Computer Game

System shared running-map diagram

Page 9: Motivate AI Class with Interactive Computer Game

CBR case representation

Page 10: Motivate AI Class with Interactive Computer Game

Similarity function

Similarity (T, S ) =

f (Ti, Si ) = (1 - |Ti – Si | / 15) * 4

25

1

( , )i

f Ti Si

Page 11: Motivate AI Class with Interactive Computer Game

Case Acquisition With 5x5 individual perceptions, the

system will have only 5x5x4x4 equals to 400 different results

If the system only adopts 100 cases then the system may only give the CBR agent ¼ chances to find the right move

the study decides to set up an 85% similarity of learning threshold while the study doing the 100 case training

Page 12: Motivate AI Class with Interactive Computer Game

Case Acquisition

For helping CBR agent find better suggestion, the study decides to adopt more cases with higher recognition rate in the bonus less areas

with 200 cases and over 90% learning threshold, at least, the CBR agent will have ½ chance to find the right suggestion

Page 13: Motivate AI Class with Interactive Computer Game
Page 14: Motivate AI Class with Interactive Computer Game

Case database refine

the study finds out there is an overfiting problem of 200 cases CBR agent since it did not has better choice in some critical moves

The study therefore, decides to prune 200 cases of redundancy into 165 refining cases

Page 15: Motivate AI Class with Interactive Computer Game
Page 16: Motivate AI Class with Interactive Computer Game

Experiment

Page 17: Motivate AI Class with Interactive Computer Game

Evaluation

100 cases

200 cases 200 cases

Page 18: Motivate AI Class with Interactive Computer Game

Conclusion

The study expects this temptation is just the beginning for applying computer games in teaching AI and CBR technologies in IT education

The advantage of AI project or assignment is that the experiment has to adopt human competition experience