Update README.md

Author: livey

README.md

@@ -1,5 +1,5 @@
 # scalable_maddpg
-scalable multi agent reinforcement learning. Details can be found in the report [Report](scalable-multi-agent.pdf)
+scalable multi agent reinforcement learning. Details can be found in the [Report](scalable-multi-agent.pdf)
 
 ## to do list
 - [x] tune L2, does LSTM parameters need L2 regulizer?
@@ -12,7 +12,7 @@ scalable multi agent reinforcement learning. Details can be found in the report
 - [ ] add another network for prey
 - [ ] add summary for rewards of each episode
 ## Alternative to Gym
-An alternative of Gym environment is created (env.py). The rendering implementation of the envrionment is matplot. So, it would be much easier to use.  
+An alternative of Gym environment is created (env.py). The rendering implementation of the envrionment is matplot. So, it would be much easier to use. However, you need to implement the prey policy by yourself. 
 ## Results
 Here, we have done two independent runs. In each run, from episode 1 to episodes 3x10^4, three agents were in the game. At episode 3x10^4, we added three more agents into this game. Here we show the mean Q value of all the agents in our experiments.