def cross_validation(c, data_set, kernel_func=lambda x, y: x.dot(y.T), times=10):
error_rates = []
for _ in xrange(times):
train_set, test_set = training_and_test_sets(data_set)
X_tr = numpy.array([test.features[0] for test in train_set])
Y_tr = numpy.array([test.answer for test in train_set])
X_test = numpy.array([test.features[0] for test in test_set])
Y_test = numpy.array([test.answer for test in test_set])
alpha, beta = smo2(c, X_tr, Y_tr, kernel_func=kernel_func)
rate = error_rate(*get_answers_and_actual(X_tr, Y_tr, X_test, Y_test, alpha, beta, kernel_func))
print rate
error_rates.append(rate)
return sum(error_rates) / float(len(error_rates))
def main():
data_set = get_data_set()
error_rates = []
for x in xrange(40):
print "test %d" % (x + 1)
training_set, test_set = training_and_test_sets(data_set)
omega = learn_perceptron(training_set)
real_answers = [test.answer for test in test_set]
classification_answers = [numpy.sign(omega.dot(test.features.T)) for test in test_set]
rate = error_rate(classification_answers, real_answers)
error_rates.append(rate)
print "error_rate: %f" % rate
print "precision: %f" % precision(classification_answers, real_answers)
print "recall: %f" % recall(classification_answers, real_answers)
print "error mean error: %0.5f" % (sum(error_rates) / float(len(error_rates)))
def cross_validation(C, data_set, times=10):
error_rates = []
for _ in xrange(times):
train_set, test_set = training_and_test_sets(data_set)
X_tr = np.array([test.features[0] for test in train_set])
Y_tr = np.array([1 if test.answer == 1 else 0 for test in train_set])
X_test = np.array([test.features[0] for test in test_set])
Y_test = np.array([1 if test.answer == 1 else 0 for test in test_set])
cost_function = logistic_cost_function(X_tr, Y_tr, C)
omega = np.zeros((1, len(X_tr[0]), ))
result = minimize(cost_function, omega, jac=None) # jac)
omega = result.x
rate = error_rate(*get_answers_and_actual(X_test, Y_test, omega))
print rate
error_rates.append(rate)
return sum(error_rates) / float(len(error_rates))
def cross_validation(c,
data_set,
kernel_func=lambda x, y: x.dot(y.T),
times=10):
error_rates = []
for _ in xrange(times):
train_set, test_set = training_and_test_sets(data_set)
X_tr = numpy.array([test.features[0] for test in train_set])
Y_tr = numpy.array([test.answer for test in train_set])
X_test = numpy.array([test.features[0] for test in test_set])
Y_test = numpy.array([test.answer for test in test_set])
alpha, beta = smo2(c, X_tr, Y_tr, kernel_func=kernel_func)
rate = error_rate(*get_answers_and_actual(X_tr, Y_tr, X_test, Y_test,
alpha, beta, kernel_func))
print rate
error_rates.append(rate)
return sum(error_rates) / float(len(error_rates))
def main():
data_set = get_data_set()
error_rates = []
for x in xrange(40):
print "test %d" % (x + 1)
training_set, test_set = training_and_test_sets(data_set)
omega = learn_perceptron(training_set)
real_answers = [test.answer for test in test_set]
classification_answers = [
numpy.sign(omega.dot(test.features.T)) for test in test_set
]
rate = error_rate(classification_answers, real_answers)
error_rates.append(rate)
print "error_rate: %f" % rate
print "precision: %f" % precision(classification_answers, real_answers)
print "recall: %f" % recall(classification_answers, real_answers)
print "error mean error: %0.5f" % (sum(error_rates) /
float(len(error_rates)))
def cross_validation(C, data_set, times=10):
error_rates = []
for _ in xrange(times):
train_set, test_set = training_and_test_sets(data_set)
X_tr = np.array([test.features[0] for test in train_set])
Y_tr = np.array([1 if test.answer == 1 else 0 for test in train_set])
X_test = np.array([test.features[0] for test in test_set])
Y_test = np.array([1 if test.answer == 1 else 0 for test in test_set])
cost_function = logistic_cost_function(X_tr, Y_tr, C)
omega = np.zeros((
1,
len(X_tr[0]),
))
result = minimize(cost_function, omega, jac=None) # jac)
omega = result.x
rate = error_rate(*get_answers_and_actual(X_test, Y_test, omega))
print rate
error_rates.append(rate)
return sum(error_rates) / float(len(error_rates))
