Deep Learning Quick Reference

Memory

Keras-RL provides us with a class called rl.memory.SequentialMemory that provides a fast and efficient data structure that we can store the agent's experiences in:

memory = SequentialMemory(limit=50000, window_length=1)

We need to specify a maximum size for this memory object, which is a hyperparameter. As new experiences are added to this memory and it becomes full, old experiences are forgotten.