Пример #1
0
 def newEpisode(self):
     """ modify the original controller parameters and store the deltas """
     DirectAgent.newEpisode(self)
     
     if not self.evaluation:
         self.deltaList.append(self.deltas.copy())  
         self.deltas = random.uniform(-self.epsilon, self.epsilon, len(self.controller.parameters))
         self.controller.parameters = self.storedParameters + self.deltas
         self.epsilon *= self.decay