modify on readme

README.md

@@ -3,18 +3,18 @@ Referring to pymarl, qmix is implemented clearly with RNN to cope with SMAC envi
 This clear implementation can help you figure out how does QMIX work  
 
 ## Run
-Note: --optimizer=0/1 means that optimizer both `Adam` and `RMSprop` are good on this scenario, please just select one of both when running
+Note: --optimizer=0/1(N) means that optimizer both `0: Adam` and `1: RMSprop` are good on this scenario, while number N in brackets means which optimizer is more suitable. Please just select one from 0/1 when running.
 ```shell
-python main.py --map-name=3s5z --optimizer=0/1
+python main.py --map-name=3s5z --optimizer=0/1(1)
 ```
 ```shell
-python main.py --map-name=1c3s5z --optimizer=0/1
+python main.py --map-name=1c3s5z --optimizer=0/1(0)
 ```
 ```shell
-python main.py --map-name=2s3z --optimizer=0/1
+python main.py --map-name=2s3z --optimizer=0/1(1)
 ```
 ```shell
-python main.py --map-name=8m --optimizer=0/1
+python main.py --map-name=8m --optimizer=0/1(0)
 ```
 ```shell
 python main.py --map-name=2s_vs_1sc --optimizer=0
@@ -26,6 +26,7 @@ python main.py --map-name=3m --optimizer=0
 python main.py --map-name=10m_vs_11m --optimizer=0
 ```
 
+And I find that in most scenarios, `0: Adam` converge faster than `1: RMSprop`.
 ## TODO
 Now this code can do very good on part of easy scenarios like 1c3s5z, 2s3z, 3s5z and 8m, 
 and relative good on easy scenarios like 2s_vs_1sc and 3m,
@@ -52,13 +53,4 @@ I'm trying to approach the result of pymarl. At the same time, I'm also trying t
   volume = {abs/1902.04043},
   year = {2019},
 }
-```
-```tex
-@article{samvelyan19smac,
-  title = {{The} {StarCraft} {Multi}-{Agent} {Challenge}},
-  author = {Mikayel Samvelyan and Tabish Rashid and Christian Schroeder de Witt and Gregory Farquhar and Nantas Nardelli and Tim G. J. Rudner and Chia-Man Hung and Philiph H. S. Torr and Jakob Foerster and Shimon Whiteson},
-  journal = {CoRR},
-  volume = {abs/1902.04043},
-  year = {2019},
-}
 ```