def run_problem1():
for n in nums:
training_set = rdm.sample(training, n)
predicted_lables.append(train_and_predict(training_set, validation, digit_specific_collecting)[1])
error_rates = [ph.benchmark(predicted_lables[i],true_labels) for i in range(len(nums))]
plt.plot(nums, error_rates)
plt.xlabel("Number of Training Data")
plt.ylabel('Error Rate')
plt.title('Number of Training Data vs. Error Rate')
def run_problem1():
for n in nums:
training_set = rdm.sample(training, n)
predicted_lables.append(
train_and_predict(training_set, validation,
digit_specific_collecting)[1])
error_rates = [
ph.benchmark(predicted_lables[i], true_labels)
for i in range(len(nums))
]
plt.plot(nums, error_rates)
plt.xlabel("Number of Training Data")
plt.ylabel('Error Rate')
plt.title('Number of Training Data vs. Error Rate')
def run_k_folds(k_folds, C=1, specific_collecting=digit_specific_collecting, specific_true_labels=digit_specific_true_labels, dataset=digits, train_labels="train_labels", train_objects="train_images", k=10):
error_rate_k_folds = 0
for i in range(k):
validation_fold = k_folds[i]
training_folds = []
for j in range(k):
if j!=i:
training_folds+=k_folds[j]
svc, predicted_labels_k_folds = train_and_predict(training_folds, validation_fold, specific_collecting, C, dataset, train_labels, train_objects)
true_labels_k_folds = specific_true_labels(validation_fold)
error_rate = ph.benchmark(predicted_labels_k_folds, true_labels_k_folds)
print "Error Rate for C=",C," is ", error_rate
error_rate_k_folds += error_rate
return svc, error_rate_k_folds/float(k)
def run_k_folds(k_folds,
C=1,
specific_collecting=digit_specific_collecting,
specific_true_labels=digit_specific_true_labels,
dataset=digits,
train_labels="train_labels",
train_objects="train_images",
k=10):
error_rate_k_folds = 0
for i in range(k):
validation_fold = k_folds[i]
training_folds = []
for j in range(k):
if j != i:
training_folds += k_folds[j]
svc, predicted_labels_k_folds = train_and_predict(
training_folds, validation_fold, specific_collecting, C, dataset,
train_labels, train_objects)
true_labels_k_folds = specific_true_labels(validation_fold)
error_rate = ph.benchmark(predicted_labels_k_folds,
true_labels_k_folds)
print "Error Rate for C=", C, " is ", error_rate
error_rate_k_folds += error_rate
return svc, error_rate_k_folds / float(k)
