diff options
Diffstat (limited to 'qlearning.py')
| -rwxr-xr-x | qlearning.py | 14 |
1 files changed, 8 insertions, 6 deletions
diff --git a/qlearning.py b/qlearning.py index 63ce70c..c731253 100755 --- a/qlearning.py +++ b/qlearning.py @@ -1,5 +1,5 @@ #!/usr/bin/env python -import sys,random,os +import sys,random,os,statistics import numpy as np # Import snake game @@ -119,7 +119,7 @@ class QTable: generation+=self.save_every with open(self.file+"_generation","w") as f: f.write(str(generation)) - print("Checkpointing generation "+str(generation)) + print("----------------------------- Checkpointing generation "+str(generation)) self.save_counter=0 def get_action(self,state): @@ -140,8 +140,9 @@ class QTable: # Perform learning -width,height=10,10 +width,height=40,30 # My advice is start with a small grid 5x5 to have many interaction and avoid early toy effect perf=0 +perf_list=list() last_state=None last_action=None game=Snake(length=1,fps=500,startat=(random.randint(0,width-1),random.randint(0,height-1)),grid_width=width,grid_height=height) @@ -172,8 +173,8 @@ while True: if stuck>=(game.grid_width*game.grid_height)/stuck_tolerance: stuck=0 stuck_count+=1 - action=qtable.get_random_action() - print("Stuck! Try a random action...") + game.new_apple() + print("Stuck! Try with a new apple...") if stuck_count>2: print("Can't get out of stuck. Abort!") stuck_count=0 @@ -187,5 +188,6 @@ while True: # Measurements score=game.last_score + perf_list.append(score) perf=max(perf,score) - print("Game ended with "+str(score)+" best so far is "+str(perf))
\ No newline at end of file + print("Game ended with "+str(score)+" best so far is "+str(perf)+ " median is "+str(statistics.median(perf_list)))
\ No newline at end of file |