changed 5 years ago  
Linked with GitHub

基於遊戲的機器學習入門Syllabus

tags: 基於遊戲的機器學習入門

Time : 15:10~18:00 Thursday
Loction : 資訊新館 65304
Lecturer : 蘇文鈺

Adjunct Expert : 陳麒麟
Teaching Assistant : 陳如琰、丁俊哲、戴逸任

Course Materials

  1. Introduction to machine learning textbook
  2. Online Game Based combat platform
  3. Online ppt
  4. Online teaching videos
  5. Reference: Stanford CS229

Target

  1. Learn basic machine learning algorithms
  2. Use machine learning tools
  3. Understand data and its preprocessing
  4. Training data and its annotation is the key
  5. Learning by doing and gaming.

1st quarter

Purpose: Get familiar with the course requirement, the offline game platform and the online game platform. Collect and annotate the data.

week 1 : Introduction of the course and applications using ML. How this course will proceed this semester. Especially in the COVID-19 period.

week 2 : Setup of Python programming environment. Use of Github.

week 3 3/19) : Build and Practice of the offline Game Platform and the online Game Platform.

week 4 (3/26) : Practice using the online Game Platform. Intro. to basic classifiers

2nd quarter

Purpose: Application of simple machine learning tools to a computer game, called Breakout Clone, also called Arkanoid. Object positions provided.

week 5 (4/2) : 兒童節、民族掃墓節補假。

Week 6 (4/9): Introduction、K-means ans KNN

Week 7 (4/23): 1st online contest, New Breakout Clone (Behavior is a little bit different!) (10%)

Week 8: Support Vector Machine

3rd quarter

Purpose: Application of more machine learning tools to New Pong and Racing Cars.

Week 9: Machin Learning Basics and Tips

Week 10: Neural Network

Week 11: 2nd online contest. New Pong (15%)

Week 12: test of statistical hypothesis

Final quarter

Purpose: Application of machine learning tools to Racing Cars 2. Object positions provided.

Week 13: Decision tree and Random Forest

Week 14: reinforcement learning(I)

Week 15: reinforcement learning(II)

Week 16: 3rd contest. Racing Cars 1. (20%)

Week 17: reinforcement learning, 端午節(放假)。

Week 18: Final contest (40%)(6/29) and final report (5%)

Grading :

  1. contest 1 : 10% + report 2%
  2. contest 2 : 15% + report 3%
  3. contest 3 : 20% + report 5%
  4. contest 4 : 40% + report 5%
  5. Bonus : 10%
  6. Contest 1: You get 5% if joining the contest. You get full grade when you pass the required levels. If you cannot pass this contest, you should consider to drop this course.
  7. Contest 2: You get 5% if joining the contest. You get full grade(15%) when your program and/or model defeats the TA’s. Those who couldn’t defeat TA’s will get the grades depending on the time your model stays in the game.
  8. Contest 3: You get 10% if joining the contest. The highest position gets full grade. The others get grades ranging from 11% to 19%. 6 cars in a game. The winner of a game joins the next game until the championship game.
  9. Contest 4: You get 20% if joining the contest. The highest position gets full grade. The others get grades ranging from 21% to 39%. 6 cars in a game. The winner of a game joins the next game until the championship game.
  10. Note 1: If you hand in all reports and join all contests, you may get 15+10+7+10+20=52% at least. If you also hand the bonus HW, you will certainly pass the course.
  11. Grade calculation of Contest 3 and 4:
    Winner of each contest with Rank=0. The second position with Rank=1, etc.
    Calculation:TBD.
Select a repo