/Atmospheric-Blocking-Classification-Persistent-Homology-2D-Convolutoional-Neural-Network

Atmospheric Blocking Classification: Persistent Homology + 2D Convolutoional Neural Network

Primary LanguagePython

INSTRUCTION:

source mySetup_nersc.source

or 

module load tensorflow/intel-1.12.0-py36

1)====== prepare train/val/test data set
mkdir out data
./format_Func2class.py

It will result with simular data split, files stpred in data/

split for dom= train X (9250, 784)
     train Y balance:0.523
split for dom= val X (1309, 784)
     val Y balance:0.514
split for dom= test X (2630, 784)
     test Y balance:0.519

2) ======== run the training over 10 epochs
./train_Func2class.py -e 5

There are 2 designes of ML avaliable:
 --design cnn (default)
 --design lstm  (runs 100 slower)

3)= = = = =   Predicting based on trained model
 ./predict_Func2class.py