def na_2_model():
#set up nnet params
layer_list=[30]
epsilon_init=0.12
alpha=0.02
n_iter=5000
lambd=1
threshold=0.5
seed=1234
cop_layer_list=copy.deepcopy(layer_list)
#read data
train,test=read_data()
#get train,test pre pre processed :D
train,test,y_matrix,id_matrix=pre_process_(train,test)
#get train,test and indexes with age
train_with_age,test_with_age,age_train_idxs,age_test_idxs=pre_process_2(train,test)
other=np.setdiff1d(test.index.values,age_test_idxs)
#get train,test without age
train_without_age,test_without_age=pre_process_3(train,test,age_test_idxs)
theta=nnet(train_with_age,y_matrix[age_train_idxs],layer_list,alpha,epsilon_init,n_iter,lambd,seed)
pred1=predict_nnet(teTitanicst_with_age,y_matrix[age_test_idxs],threshold,layer_list,theta,epsilon_init,seed)
theta=nnet(train_without_age,y_matrix,layer_list,alpha,epsilon_init,n_iter,lambd,seed)
pred2=predict_nnet(test_without_age,y_matrix[other],threshold,layer_list,theta,epsilon_init,seed)
pred=test.index.values.astype('float')
pred[age_test_idxs]=pred1
pred[other]=pred2
pred=pred.reshape((len(pred),1))
#merge PassengerId and predictions to create the submission DataFrame
df=pd.DataFrame(np.concatenate([id_matrix,pred],axis=1))
#set up column names
df.columns=['PassengerId','Survived']
#turn PassengerId column from float to int
df['PassengerId']=df['PassengerId'].astype('int')
df['Survived']=(df['Survived']>threshold)*1
#create submission csv
df.to_csv(filename,header=True,index=False)
def submit(filename):
"""(string) -> None
Runs perceptron on the titanic dataset and generates
a submission file 'filename'.csv
"""
#read data
train, test = read_data()
#does some of the trevor stephens pre-processing, check docstring of function
train, test, y_matrix, id_matrix = pre_process_(train, test)
#turns dataframes into matrices so the logistic regression code doesnt flip out
train_, test_ = turn_into_matrices(train, test)
#set nnet params
layer_list = [10]
epsilon_init = 0.12
alpha = 0.1
n_iter = 1000
lambd = 0
threshold = 0.5
seed = 1234
#get theta
theta = nnet(train_, y_matrix, layer_list, alpha, epsilon_init, n_iter,
lambd, seed)
#get predictions in the test set
pred = predict_nnet(test_, y_matrix, threshold, layer_list, theta,
epsilon_init, seed)
#merge PassengerId and predictions to create the submission DataFrame
df = pd.DataFrame(np.concatenate([id_matrix, pred], axis=1))
#set up column names
df.columns = ['PassengerId', 'Survived']
#turn PassengerId column from float to int
df['PassengerId'] = df['PassengerId'].astype('int')
#turn predictions from float into binary int 0,1
df['Survived'] = (df['Survived'] > threshold) * 1
#create submission csv
df.to_csv(filename, header=True, index=False)
def imputation(train,test,y):
# #set up logistic regression parameters
# n_iter=10000
# alpha=0.1
# #get theta
# _,theta=linear_regression(train,y,alpha,n_iter)
# m=np.shape(test)[0]
# test=normalize_features(test)
# test=np.concatenate((np.ones((m,1)),test),axis=1)
# #get predictions in the test set
# pred=hyp_lin_r(test,theta)
# #train=np.concatenate([train,pred],axis=1)
layer_list=[20]
epsilon_init=0.12
alpha=0.1
n_iter=1000
lambd=0
threshold=0.5
seed=1234
theta=nnet(train,y,layer_list,alpha,epsilon_init,n_iter,lambd,seed)
layer_list=[20]
pred=predict_nnet(test,y,threshold,layer_list,theta,epsilon_init,seed)
return pred
def submit(filename):
"""(string) -> None
Runs perceptron on the titanic dataset and generates
a submission file 'filename'.csv
"""
#read data
train,test=read_data()
#does some of the trevor stephens pre-processing, check docstring of function
train,test,y_matrix,id_matrix=pre_process_(train,test)
#turns dataframes into matrices so the logistic regression code doesnt flip out
train_,test_=turn_into_matrices(train,test)
#set nnet params
layer_list=[10]
epsilon_init=0.12
alpha=0.1
n_iter=1000
lambd=0
threshold=0.5
seed=1234
#get theta
theta=nnet(train_,y_matrix,layer_list,alpha,epsilon_init,n_iter,lambd,seed)
#get predictions in the test set
pred=predict_nnet(test_,y_matrix,threshold,layer_list,theta,epsilon_init,seed)
#merge PassengerId and predictions to create the submission DataFrame
df=pd.DataFrame(np.concatenate([id_matrix,pred],axis=1))
#set up column names
df.columns=['PassengerId','Survived']
#turn PassengerId column from float to int
df['PassengerId']=df['PassengerId'].astype('int')
#turn predictions from float into binary int 0,1
df['Survived']=(df['Survived']>threshold)*1
#create submission csv
df.to_csv(filename,header=True,index=False)
for I in range(1, 5):
file_name = cifar_dir + '/data_batch_' + str(I + 1)
file_id = open(file_name, 'rb')
dict = cPickle.load(file_id)
file_id.close()
data = np.concatenate((data, dict['data']), axis=0)
labels = np.concatenate((labels, dict['labels']), axis=0)
return data, labels
nnet_size = np.array([[992, 500], [500, 500], [500, 10]])
learning_rate = 0.1
num_layers = 3
mlp = nnet(nnet_size, num_layers, learning_rate)
epoch = 3
cifar_dir = '/home/habtegebrial/Desktop/python-destin/cifar-10-batches-py/'
trainData = np.array([])
for I in range(0, 50000, 1):
#for I in range(499, 500, 500):
Name = '/home/habtegebrial/GSoc/python-destin/testing/train/' + str(I + 1) + '.txt'
file_id = open(Name, 'r')
#print np.shape(np.ravel(np.loadtxt(Name)))
Temp = np.array(cPickle.load(file_id))
print np.shape(Temp)
file_id.close()
trainData = np.hstack((trainData, Temp))
