1. # of pions 1->1 : multi->1 (original MC->recon) = ~ 50:50

2. Tried with missing mass cut got ~70% of purity.

3. Tried to apply kinds of neural network with parameters as bellow but just got few more % purity.

1. W

2. Electron momentum

3. Q2

4. Electron Angle

5. Missing Mass Squared

6. Pion Momentum

7. Pion Angle