def exampleUsage():
'''This is just here to show how to run the models'''
training_data = data.loadTraining()
out = gaussianNBayes(training_data['xlabels'], training_data['ylabels'], \
training_data['xlabels'])
out1 = ridgeRegression(training_data['xlabels'], training_data['ylabels'], \
training_data['xlabels'], 3)
out2 = randomForest(training_data['xlabels'], training_data['ylabels'], \
training_data['xlabels'])
import pkg_resources
import sys
import os
import sklearn
import numpy as np
from sklearn.neural_network import MLPClassifier
from sklearn.grid_search import GridSearchCV
from sklearn import cross_validation
import dataIO as data
import datetime
from grid_search import grid_search
import time
full_train = data.loadTraining()
x_train = full_train['xlabels']
y_train = full_train['ylabels']
full_test = data.loadTest()
x_test = full_test['xlabels']
# the 2 values for alpha are needed because there needs to be 2 parameters for
# the grid search. Could just remove the grid search
parameters = {'hidden_layer_sizes': [tuple([10] * 1)],\
'activation': ['logistic'], \
'learning_rate': ['constant'], \
'alpha': [7.0] * 2
}
num_folds = np.prod(np.array([len(parameters[key]) for key in parameters]))
print "Number of folds: " + str(num_folds)
'rf_regress_train.csv',
'rf_regress_train_2.csv',
'svm_regress_train.csv'
]
testDataFiles = ['ada_regress_test.csv',
'gnb_class_test.csv',
'knn_regress_test.csv',
'log_regress_test.csv',
'nn_class_test.csv',
'rf_regress_test.csv',
'rf_regress_test_2.csv',
'svm_regress_test.csv'
]
y_train = data.loadTraining()['ylabels']
x_train = np.zeros((len(trainingDataFiles), len(y_train)))
x_train = np.transpose([loadModelOut(model) for model in trainingDataFiles])
x_test = np.zeros((len(testDataFiles), len(y_train)))
x_test = np.transpose([loadModelOut(model) for model in testDataFiles])
in_layer = len(trainingDataFiles)
parameters = {'hidden_layer_sizes': ((in_layer,), (in_layer, in_layer/2), (in_layer, 2*in_layer/3, in_layer/3)),
'alpha': np.logspace(-5, -3, 15),
'learning_rate': ('constant', 'invscaling')
}
kf_total = cross_validation.KFold(len(x_train), n_folds=10,\