def main(args):
lr = args.lr
n_hu = args.hu
epoch = args.epoch
train = args.train
test = args.test
test = test.replace('\r','') #Removes the carriage return cuz I use windows
data_origin = test.replace('_test.json','')
data_origin = data_origin.replace('data/','')
#Load the data in and convert to numpy
with open(train,"r") as read_file:
train = json.load(read_file)
train = json2numpy(train)
train = train.train_pp()
with open(test,"r") as read_file:
test = json.load(read_file)
test = json2numpy(test)
test = test.test_pp(train)
xlist = []
trainlist = []
testlist = []
if n_hu > 0:
title = 'Neural Network: F1 vs. Epochs'
filename = 'nnet_f1_curve.png'
caption = ('Dataset: ' + data_origin + ', Learning rate: ' + str(lr) + ' , Number of hidden units: ' + str(n_hu) + ', Maximum number of epochs: ' + str(epoch))
for e in range(1,epoch+1):
xlist.append(e)
x = 'nn'
w = nn_train(lr,n_hu,e,train, screen_print = 0)
po = nn_predict(test,w, screen_print = 0)
testlist.append(po[1])
po = nn_predict(train,w, screen_print = 0)
trainlist.append(po[1])
else:
title = 'Logistic Regression: F1 vs. Epochs'
filename = 'logistic_f1_curve.png'
caption = ('Dataset: ' + data_origin + ', Learning rate: ' + str(lr) + ', Maximum number of epochs: ' + str(epoch))
for e in range(1,epoch+1):
xlist.append(e)
x = 'lr'
w = lr_train(lr,e, train, screen_print = 0)
po = lr_predict(test,w, screen_print = 0)
testlist.append(po[1])
po = lr_predict(train,w, screen_print = 0)
trainlist.append(po[1])
#Plot the data
plt.plot(xlist,trainlist)
plt.plot(xlist,testlist)
plt.legend(['Train','Test'], loc = "lower right")
plt.xlabel('Number of epochs')
plt.ylabel('F1')
plt.ylim([0.0, 1.05])
plt.xlim([0.0, epoch+5])
plt.title(caption, fontsize = 9, wrap = True)
plt.suptitle(title)
plt.savefig('./figs/'+filename)
def main(args):
method = args.method
trees = args.trees
depth = args.depth
train = args.train
test = args.test
test = test.replace('\r','') #Removes the carriage return cuz I use windows
#Load the data in and convert to numpy
with open(train,"r") as read_file:
train = json.load(read_file)
train = json2numpy(train)
with open(test,"r") as read_file:
test = json.load(read_file)
test = json2numpy(test)
#Pull the true labels
truth = test.labels
#Generate the predictions from the ensemble
if method == 'bag':
predictions = bootstrap(trees, depth, train, test, display = False)
elif method == 'boost':
predictions = adaboost(trees, depth, train, test, display = False)
else:
print('Invalid ensemble method.')
build_confusion_matrix(test, predictions, verbose = True)
def main(args):
trees = args.trees
depth = args.depth
train = args.train
test = args.test
test = test.replace('\r','') #Removes the carriage return cuz I use windows
#Load the data in and convert to numpy
with open(train,"r") as read_file:
train = json.load(read_file)
train = json2numpy(train)
with open(test,"r") as read_file:
test = json.load(read_file)
test = json2numpy(test)
bootstrap(trees, depth, train, test, display = True)
def main(args):
lr = args.lr
epoch = args.epoch
train = args.train
test = args.test
test = test.replace('\r','') #Removes the carriage return cuz I use windows
#Load the data in and convert to numpy
with open(train,"r") as read_file:
train = json.load(read_file)
train = json2numpy(train)
train = train.train_pp()
with open(test,"r") as read_file:
test = json.load(read_file)
test = json2numpy(test)
test = test.test_pp(train)
#lr_train trains the weights for the logistic regression
w = lr_train(lr,epoch,train)
#lr_predict predicts classes for test set
predictions = lr_predict(test,w)
def main(args):
method = args.method
trees = args.trees
depth = args.depth
train = args.train
test = args.test
test = test.replace('\r','') #Removes the carriage return cuz I use windows
data_origin = test.replace('_test.json','')
data_origin = data_origin.replace('data/','')
#Load the data in and convert to numpy
with open(train,"r") as read_file:
train = json.load(read_file)
train = json2numpy(train)
with open(test,"r") as read_file:
test = json.load(read_file)
test = json2numpy(test)
#Pull the true labels
truth = test.labels
#Generate the predictions from the ensemble, create list of coordinates
xlist = []
ylist = []
if method == 'bag':
title = 'Decision Tree Bagging: Accuracy vs. Number of trees'
filename = 'bagged_tree_plot.pdf'
caption = ('Dataset: ' + data_origin)
depth1 = depth - 2
depth2 = depth
depth3 = depth + 2
for j in [depth1,depth2,depth3]:
for i in range(1,trees+1):
predictions = bootstrap(i, j, train, test, display = False)
accuracy = sum((truth == predictions).astype(float)) / test.length
xlist.append(i)
ylist.append(accuracy)
plt.plot(xlist,ylist)
xlist = []
ylist = []
elif method == 'boost':
title = 'Decision Tree AdaBoost: Accuracy vs. Number of trees'
filename = 'boosted_tree_plot.pdf'
caption = ('Dataset: ' + data_origin)
depth1 = depth - 1
depth2 = depth
depth3 = depth + 1
for j in [depth1,depth2,depth3]:
for i in range(1,trees+1):
predictions = adaboost(i, j, train, test, display = False)
accuracy = sum((truth == predictions).astype(float)) / test.length
xlist.append(i)
ylist.append(accuracy)
plt.plot(xlist,ylist)
xlist = []
ylist = []
else:
print('Invalid ensemble method.')
#Plot the data
plt.legend([depth1,depth2,depth3], loc = "lower right", title = "Maximum tree depth", fancybox = True)
plt.xlabel('Ensemble size')
plt.ylabel('Accuracy')
plt.ylim([0.0, 1.05])
plt.xlim([1, trees])
plt.title(caption, fontsize = 9, wrap = True)
plt.suptitle(title)
plt.savefig(filename)