def load_data(self):
training_filepath = self.training_file
testing_filepath = self.testing_file
if not self.split and not self.bulk:
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath, n_features=self.n_features)
testing_loader = LoadLibsvm.LoadLibSVM(filename=testing_filepath, n_features=self.n_features)
self.x_training, self.y_training = training_loader.load_all_data()
self.x_testing, self.y_testing = testing_loader.load_all_data()
if self.split and not self.bulk:
Print.Print.result2("Data Splitting ...")
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath, n_features=self.n_features)
x_all, y_all = training_loader.load_all_data()
ratio = 0.8
size = len(x_all)
split_index = int(size * ratio)
self.x_training = x_all[:split_index]
self.x_testing = x_all[split_index:]
self.y_training = y_all[:split_index]
self.y_testing = y_all[split_index:]
if (self.labelfix):
self.y_training = self.y_training[self.y_training == 2] = -1
self.y_testing = self.y_testing[self.y_testing == 2] = -1
if self.bulk and not self.split:
# here we consider a larger testing data set.
print("Loading Bulk Training File")
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath, n_features=self.n_features)
self.x_training, self.y_training = training_loader.load_all_data()
return self.x_training, self.y_training
else:
return self.x_training, self.y_training, self.x_testing, self.y_testing
print("Training and Testing data loaded ...")
def load_light_data(training_file, testing_file, split=False):
x_training = []
x_testing = []
y_training = []
y_testing = []
training_filepath = training_file
testing_filepath = testing_file
if not split:
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath, n_features=features)
testing_loader = LoadLibsvm.LoadLibSVM(filename=testing_filepath, n_features=features)
x_training, y_training = training_loader.load_all_data()
x_testing, y_testing = testing_loader.load_all_data()
if split:
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath, n_features=features)
x_all, y_all = training_loader.load_all_data()
ratio = 0.8
size = len(x_all)
split_index = int(size * ratio)
x_training = x_all[:split_index]
x_testing = x_all[split_index:]
y_training = y_all[:split_index]
y_testing = y_all[split_index:]
print("Training and Testing data loaded ...")
return x_training, y_training, x_testing, y_testing
def load_data(self):
training_filepath = self.training_file
testing_filepath = self.testing_file
if not self.split:
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath,
n_features=self.n_features)
testing_loader = LoadLibsvm.LoadLibSVM(filename=testing_filepath,
n_features=self.n_features)
self.x_training, self.y_training = training_loader.load_all_data()
self.x_testing, self.y_testing = testing_loader.load_all_data()
if self.split:
Print.Print.result2("Data Splitting ...")
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath,
n_features=self.n_features)
x_all, y_all = training_loader.load_all_data()
ratio = 0.8
size = len(x_all)
split_index = int(size * ratio)
self.x_training = x_all[:split_index]
self.x_testing = x_all[split_index:]
self.y_training = y_all[:split_index]
self.y_testing = y_all[split_index:]
if (self.labelfix):
self.y_training = self.y_training[self.y_training == 2] = -1
self.y_testing = self.y_testing[self.y_testing == 2] = -1
return self.x_training, self.y_training, self.x_testing, self.y_testing
def load_data(training_file, testing_file, split=False, n_features=10):
training_filepath = training_file
testing_filepath = testing_file
if not split:
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath,
n_features=n_features)
testing_loader = LoadLibsvm.LoadLibSVM(filename=testing_filepath,
n_features=n_features)
x_training, y_training = training_loader.load_all_data()
x_testing, y_testing = testing_loader.load_all_data()
if split:
print("Data Splitting ...")
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath,
n_features=n_features)
x_all, y_all = training_loader.load_all_data()
ratio = 0.6
size = len(x_all)
split_index = int(size * ratio)
x_training = x_all[:split_index]
x_testing = x_all[split_index:]
y_training = y_all[:split_index]
y_testing = y_all[split_index:]
return x_training, y_training, x_testing, y_testing
def load_svm_data(self):
training_filepath = self.training_file
testing_filepath = self.testing_file
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath, n_features=22)
testing_loader = LoadLibsvm.LoadLibSVM(filename=testing_filepath, n_features=22)
x_training, y_training = training_loader.load_all_data()
x_testing, y_testing = testing_loader.load_all_data()
return x_training, y_training, x_testing, y_testing
def advanced_test(self, x_testing, y_testing, w):
#y_pred = self.svm.predict(X=self.x_testing)
#self.acc = self.svm.get_accuracy(y_test=self.y_testing, y_pred=y_pred)
self.acc = 0
if self.bulk:
testing_filepath = self.testing_file
print("Loading Bulk Testing Files")
files = os.listdir(testing_filepath)
print("File Path : " + testing_filepath)
print(files)
self.bulk_testing_x = []
self.bulk_testing_y = []
self.acc = 0
for file in files:
print("Loading Testing Bulk File : " + file)
testing_loader = LoadLibsvm.LoadLibSVM(
filename=testing_filepath + "/" + file,
n_features=self.n_features)
x_testing, y_testing = testing_loader.load_all_data()
y_pred = self.svm.custom_predict(x_testing, w=w)
self.acc += self.svm.get_accuracy(y_test=y_testing,
y_pred=y_pred)
self.acc = self.acc / len(files)
else:
y_pred = self.svm.custom_predict(X=x_testing, w=w)
self.acc = self.svm.get_accuracy(y_test=y_testing, y_pred=y_pred)
return self.acc
def load_data(self):
Print.Print.info1("Loading Data")
training_filepath = self.training_file
testing_filepath = self.testing_file
if not self.split and not self.bulk:
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath,
n_features=self.n_features)
testing_loader = LoadLibsvm.LoadLibSVM(filename=testing_filepath,
n_features=self.n_features)
self.x_training, self.y_training = training_loader.load_all_data()
self.x_testing, self.y_testing = testing_loader.load_all_data()
if self.split and not self.bulk:
Print.Print.info1("Splitting data ...")
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath,
n_features=self.n_features)
x_all, y_all = training_loader.load_all_data()
ratio = 0.8
size = len(x_all)
split_index = int(size * ratio)
self.x_training = x_all[:split_index]
self.x_testing = x_all[split_index:]
self.y_training = y_all[:split_index]
self.y_testing = y_all[split_index:]
if (self.labelfix):
# this logic can varied depending on the target labels
# TODO : Change the logic
self.y_training = self.y_training[self.y_training == 2] = -1
self.y_testing = self.y_testing[self.y_testing == 2] = -1
if self.bulk and not self.split:
# here we consider a larger testing data set.
# here we consider a larger testing data set.
print("Loading Bulk Training File")
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath,
n_features=self.n_features)
self.x_training, self.y_training = training_loader.load_all_data()
return self.x_training, self.y_training
else:
return self.x_training, self.y_training, self.x_testing, self.y_testing
print("Training and Testing data loaded ...")
def load_training_data(training_file, split=False):
x_training = []
y_training = []
training_filepath = training_file
if not split:
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath, n_features=features)
x_training, y_training = training_loader.load_all_data()
print("Training data loaded ...")
return x_training, y_training
def load_random_test_data(testing_filepath):
print("Loading Bulk Testing Files")
files = os.listdir(testing_filepath)
print("File Path : " + testing_filepath)
print(files)
bulk_testing_x = []
bulk_testing_y = []
file_size = len(files)
random_id = np.random.randint(0,file_size)
print("Loading Testing Bulk File : " + files[random_id])
testing_loader = LoadLibsvm.LoadLibSVM(filename=testing_filepath + "/" + files[random_id],
n_features=1)
x_testing, y_testing = testing_loader.load_all_data()
return x_testing, y_testing
def case3():
training_filepath = "/home/vibhatha/data/svm/ijcnn1/ijcnn1_training"
testing_filepath = "/home/vibhatha/data/svm/ijcnn1/ijcnn1_testing"
loader = LoadLibsvm.LoadLibSVM(filename=training_filepath, n_features=22)
x_training, y_training, x_testing, y_testing = loader.load_data()
svm = SVM.SVM(X=x_training, y=y_training, alpha=1, n_features=22)
# alpha=0.0003, epochs=1000 #good
alpha = 0.1
epochs = 500
svm.train(X=x_training, y=y_training, alpha=alpha, epochs=epochs)
y_pred = svm.predict(X=x_testing)
acc = svm.get_accuracy(y_test=y_testing, y_pred=y_pred)
print("Accuracy : " + str(acc) + "%")
fp = open("logs/results.txt", "a")
fp.write("alpha : " + str(alpha) + ", epochs : " + str(epochs) +
", accuracy : " + str(acc) + "%\n")
fp.close()
def test(self):
self.acc_m = 0
self.acc_sgd = 0
if self.bulk:
testing_filepath = self.testing_file
print("Loading Bulk Testing Files")
files = os.listdir(testing_filepath)
print("File Path : " + testing_filepath)
print(files)
self.bulk_testing_x = []
self.bulk_testing_y = []
for file in files:
print("Loading Testing Bulk File : " + file)
testing_loader = LoadLibsvm.LoadLibSVM(
filename=testing_filepath + "/" + file,
n_features=self.n_features)
x_testing, y_testing = testing_loader.load_all_data()
y_pred_m = self.svm.custom_predict(x_testing, w=self.w_m)
self.acc_m += self.svm.get_accuracy(y_test=y_testing,
y_pred=y_pred_m)
y_pred_sgd = self.svm.custom_predict(x_testing, w=self.w_sgd)
self.acc_sgd += self.svm.get_accuracy(y_test=y_testing,
y_pred=y_pred_sgd)
self.acc_m = self.acc_m / len(files)
self.acc_sgd = self.acc_sgd / len(files)
else:
y_pred_m = self.svm.custom_predict(X=self.x_testing, w=self.w_m)
self.acc_m = self.svm.get_accuracy(y_test=self.y_testing,
y_pred=y_pred_m)
y_pred_sgd = self.svm.custom_predict(X=self.x_testing,
w=self.w_sgd)
self.acc_sgd = self.svm.get_accuracy(y_test=self.y_testing,
y_pred=y_pred_sgd)
return self.acc_m, self.acc_sgd
from operations import LoadLibsvm
import numpy as np
from matplotlib import pyplot as plt
X = np.array([[1, 1], [2, 1], [3, 1], [4, 1], [1, 5], [2, 6], [3, 7], [4, 5]])
y = np.array([1, 1, 1, 1, -1, -1, -1, -1])
X_test = np.array([[1, 1.25], [2.1, 1.15], [3.1, 1.45], [4.23, 1.21],
[1.3, 5.25], [2.11, 6.24], [3.3, 7.24], [4.212, 5.78]])
#plt.scatter(X[:,0],X[:,1])
#plt.show()
dataset = 'ijcnn1'
training_filepath = '/home/vibhatha/data/svm/' + dataset + '/training.csv'
testing_filepath = '/home/vibhatha/data/svm/' + dataset + '/testing.csv'
n_features = 300
split = False
training_loader = LoadLibsvm.LoadLibSVM(filename=training_filepath,
n_features=n_features)
x_training = []
y_training = []
x_testing = []
y_testing = []
if split == True:
x_all, y_all = training_loader.load_all_data()
ratio = 0.8
size = len(x_all)
split_index = int(size * ratio)
x_training = x_all[:split_index]
x_testing = x_all[split_index:]
y_training = y_all[:split_index]
y_testing = y_all[split_index:]
## iterate through all the test samples.
overall_accuracy_sgd = 0
overall_accuracy_mom = 0
overall_accuracy_ada = 0
overall_accuracy_rmsprop = 0
overall_accuracy_adam = 0
files = os.listdir(testing_file)
print("File Path : " + testing_file)
print(files)
bulk_testing_x = []
bulk_testing_y = []
file_size = len(files)
random_id = np.random.randint(0, file_size)
print("Loading Testing Bulk File : " + files[random_id])
for file in files:
testing_loader = LoadLibsvm.LoadLibSVM(filename=testing_file + "/" + file,
n_features=1)
x_testing, y_testing = testing_loader.load_all_data()
overall_accuracy_sgd += bmsvmsgd.advanced_test(x_testing=x_testing, y_testing=y_testing, w=w_final_sgd)
overall_accuracy_mom += bmsvmsgdmomentum.advanced_test(x_testing=x_testing, y_testing=y_testing, w=w_final_mom)
overall_accuracy_ada += bmsvmsgdada.advanced_test(x_testing=x_testing, y_testing=y_testing, w=w_final_ada)
overall_accuracy_rmsprop += bmsvmsgdrmsprop.advanced_test(x_testing=x_testing, y_testing=y_testing, w=w_final_rmsprop)
overall_accuracy_adam += bmsvmsgdadam.advanced_test(x_testing=x_testing, y_testing=y_testing,
w=w_final_adam)
overall_accuracy_sgd = overall_accuracy_sgd/ file_size
overall_accuracy_mom = overall_accuracy_mom / file_size
overall_accuracy_ada = overall_accuracy_ada / file_size
overall_accuracy_rmsprop = overall_accuracy_rmsprop / file_size
overall_accuracy_adam = overall_accuracy_adam / file_size
## stats
