In [1]:

    

import nn_learn


    

In [2]:

    

import pickle
dat = pickle.load(open("data/TrainingData/pickled_generated_sets",'rb'))
pca_dat = pickle.load(open("data/TrainingData/pickled_generated_sets_pca",'rb'))


    

In [3]:

    

pca_dat['train_labels'].shape


    

    
Out[3]:





(12278,)

In [4]:

    

neural_net1_pca = {
    "num_features": 12,
    "num_labels": 8,
    "num_layers": 3,
    "num_neurons": [256, 256, 256],
    "use_dropout": True,
    "model_name": "model_pca_3_layers_256",
    "learning_rate": 0.1,
    "learning_rate_decay":0.96
}


    

In [5]:

    

NN_3L_256_256_256 = nn_learn.NeuralNetwork(neural_net1_pca)


    

In [ ]:

    

NN_3L_256_256_256.train(pca_dat["train_data"], pca_dat["train_labels"],pca_dat["valid_data"],pca_dat["valid_labels"], 1000)


    

    




Training Step: 548  | total loss: 1.51852 | time: 0.984s
| SGD | epoch: 003 | loss: 1.51852 - top3: 0.7818 -- iter: 10496/12278

In [ ]:

    

pca_dat


    

In [ ]: