Create an AI agent for the single player Pong-like game.
-
Create your own training file:
Make sure to settrain = True
around lab4.py line 94 to record your states.
Play the game until you have a satisfying csv recording. -
Train:
-
Look at your csv and decide on your ML model:
Supervised/unsupervised?
Classification / regression / clustering?
https://scikit-learn.org/stable/tutorial/machine_learning_map/index.html -
Create a separate notebook/script, train.py:
Create X/y matrices: What should be your features vs output?
See 5.2/12a: X_iris/y_iris setup via pd.drop()
Train and save your ML model
See class examples or dsexample
Save your model to be used in the game, wheremodel
is the name of your scikit model:
from joblib import dump, load
dump(model, 'mymodel.joblib') #save
- Deploy:
- go back to lab4.py, set
train = False
around line 94 - load your model in the
if not train
block, around line 150:
from joblib import dump, load
model = load('mymodel.joblib') #load
- finally, complete the logic to predict the desired paddle postion/direction and to control the paddle to that location.
Here are two different types fo models from two different training sets playing solo: