This is the code repository for the paper:
Junhong Shen, Lin Yang, Theoretically Principled Deep RL Acceleration via Nearest Neighbor Function Approximation, AAAI 2021. 

To replicate the CartPole-v1 experiments, go to the CartPole folder:
1. NNAC.py: run NNAC with the internal state descriptors
2. NNAC_img_l2: run NNAC with pixel input and l2 distance metric
3. NNAC_img_learned_metric.ipybn: run NNAC with pixel input and the learned distance metric
4. baselines.py: run deep RL baselines
5. To run NEC, we use the code from https://github.com/EndingCredits/Neural-Episodic-Control.

To replicate the MuJoCo experiments, go to the MuJoCo folder:
main.py: provide the name of the algorithm, i.e., one of DDPG, TD3, NNDDPG and NNTD3, as well 
	 as the name of the environment to run the experiment