def Cross_validation(X_train, y_train):
"""交叉验证，确定超参K，同时可视化K值
:param X_train: 训练集
:param y_train: 训练标签
"""
num_folds = 5
k_choices = [1, 3, 5, 8, 10, 12, 15, 20, 50, 100]
k_accuracy = {}
# 将数据集分为5份
X_train_folds = np.array_split(X_train, num_folds)
y_train_folds = np.array_split(y_train, num_folds)
# 计算每种K值
for k in k_choices:
k_accuracy[k] = []
# 每个K值分别计算每份数据集作为测试集时的正确率
for index in range(num_folds):
# 构建数据集
X_te = X_train_folds[index]
y_te = y_train_folds[index]
X_tr = np.reshape(
X_train_folds[:index] + X_train_folds[index + 1:],
(X_train.shape[0] * (num_folds - 1) / num_folds, -1))
y_tr = np.reshape(
y_train_folds[:index] + y_train_folds[index + 1:],
(X_train.shape[0] * (num_folds - 1) / num_folds))
# 预测结果
classify = KNearestNeighbor()
classify.train(X_tr, y_tr)
y_te_pred = classify.predict(X_te, k=k)
accuracy = np.sum(y_te_pred == y_te) / float(X_te.shape[0])
k_accuracy[k].append(accuracy)
for k, accuracylist in k_accuracy.items():
for accuracy in accuracylist:
print("k = %d, accuracy = %.3f" % (k, accuracy))
# 可视化K值效果
for k in k_choices:
accuracies = k_accuracy[k]
plt.scatter([k] * len(accuracies), accuracies)
accuracies_mean = np.array(
[np.mean(v) for k, v in sorted(k_accuracy.items())])
accuracies_std = np.array(
[np.std(v) for k, v in sorted(k_accuracy.items())])
# 根据均值和方差构建误差棒图
plt.errorbar(k_choices, accuracies_mean, yerr=accuracies_std)
plt.title('Cross-validation on k')
plt.xlabel('k')
plt.ylabel('Cross-validation accuracy')
plt.show()
def test_arange(self):
train = np.arange(150).reshape(5, -1)
test = np.square(np.arange(2, 122)).reshape(4, -1)
knn = KNearestNeighbor()
knn.train(train, None)
d_two = knn.compute_distances_two_loops(test)
d_one = knn.compute_distances_one_loop(test)
d_no = knn.compute_distances_no_loops(test)
self.assertAlmostEqual(0, np.linalg.norm(d_two - d_one, ord='fro'))
self.assertAlmostEqual(0, np.linalg.norm(d_no - d_one, ord='fro'))
def test_arange(self):
train = np.arange(150).reshape(5, -1)
test = np.square(np.arange(2, 122)).reshape(4, -1)
knn = KNearestNeighbor()
knn.train(train, None)
d_two = knn.compute_distances_two_loops(test)
d_one = knn.compute_distances_one_loop(test)
d_no = knn.compute_distances_no_loops(test)
self.assertAlmostEqual(0, np.linalg.norm(d_two - d_one, ord='fro'))
self.assertAlmostEqual(0, np.linalg.norm(d_no - d_one, ord='fro'))
for i, idx in enumerate(idxs):
plt_idx = i * class_num + y + 1
plt.subplot(samples_pre_class, class_num, plt_idx)
plt.imshow(X_train[idx].astype('uint8'))
plt.axis('off')
if i == 0:
plt.title(cls)
plt.show()
X_train = X_train.reshape(500, -1)
y_train = y_train.reshape(500, -1)
X_test = X_test.reshape(10, -1)
y_test = y_test.reshape(10, -1)
classifier = KNearestNeighbor()
classifier.train(X_train, y_train)
dists = classifier.compute_distance_two_loops(X_test)
dists_one = classifier.compute_distance_one_loop(X_test)
diff = np.linalg.norm(dists - dists_one, ord='fro')
if diff < 0.001:
print('good')
else:
print('bad')
y_pred = classifier.predict_labels(dists, 1)
correct = np.where(y_pred == y_train)
print('accuracy: ', len(correct) / len(y_test))
# cross validation
num_folds = 5
print 'Test data new shape: ', X_test.shape
# Create a kNN classifier instance.
# Remember that training a kNN classifier is a noop:
# the Classifier simply remembers the data and does no further processing
# classifier = KNearestNeighbor()
# classifier.train(X_train, y_train)
# numK = [8,9,10,11,12,13,14,15,16]
numK = [12]
results = {}
bestValAcc = 0
bestK = None
for num in numK:
knn = KNearestNeighbor()
knn.train(X_train, y_train)
y_train_pred = knn.predict(X_train, k=num)
y_val_pred = knn.predict(X_val, k=num)
trainAcc = np.mean(y_train == y_train_pred)
valAcc = np.mean(y_val == y_val_pred)
print 'k: %d train accuracy: %.4f val accuracy: %.4f' % (num, trainAcc,
valAcc)
if valAcc > bestValAcc:
bestValAcc = valAcc
bestK = num
print 'best validation accuracy achieved: %.4f, with best k : %d' % (
bestValAcc, bestK)
# Based on the cross-validation results above, choose the best value for k,
y_train = y_train.reshape(-1, 1)
x_train_folds = np.array_split(x_train, num_folds)
y_train_folds = np.array_split(y_train, num_folds)
x_train_folds = np.array(x_train_folds)
y_train_folds = np.array(y_train_folds)
k_to_accuracies = {}
for k in k_choices:
k_to_accuracies.setdefault(k, [])
for i in range(num_folds):
classifier = KNearestNeighbor()
x_val_train = np.concatenate((x_train_folds[0:i], x_train_folds[i + 1:]),
axis=0)
x_val_train = x_val_train.reshape(-1, x_val_train.shape[2])
y_val_train = np.concatenate((y_train_folds[0:i], y_train_folds[i + 1:]),
axis=0)
y_val_train = y_val_train.reshape(-1, y_val_train.shape[2])
y_val_train = y_val_train[:, 0]
classifier.train(x_val_train, y_val_train)
for k in k_choices:
y_val_pred = classifier.predict_labels(x_train_folds[i], k=k)
num_correct = np.sum(y_val_pred == y_train_folds[i][:, 0])
accuracy = float(num_correct) / len(y_val_pred)
k_to_accuracies[k] = k_to_accuracies[k] + [accuracy]
num_training = 5000
mask = range(num_training)
X_train = X_train[mask]
y_train = y_train[mask]
num_test = 500
mask = range(num_test)
X_test = X_test[mask]
y_test = y_test[mask]
X_train = np.reshape(X_train, (X_train.shape[0], -1))
X_test = np.reshape(X_test, (X_test.shape[0], -1))
print(X_train.shape, X_test.shape)
classifier = KNearestNeighbor()
classifier.train(X_train, y_train)
dists = classifier.compute_distances_no_loops(X_test)
print(dists.shape)
plt.imshow(dists, interpolation='none')
plt.show()
y_test_pred = classifier.predict_labels(dists, k=5)
# Compute and print the fraction of correctly predicted examples
num_correct = np.sum(y_test_pred == y_test)
accuracy = float(num_correct) / num_test
print('Got %d / %d correct => accuracy: %f' %
(num_correct, num_test, accuracy))
X_train = X_train[mask] y_train = y_train[mask] num_test = 500 mask = range(num_test) X_test = X_test[mask] y_test = y_test[mask] # reshape训练和测试数据，转换为行的形式 X_train = np.reshape(X_train, (X_train.shape[0], -1)) X_test = np.reshape(X_test, (X_test.shape[0], -1)) print(X_train.shape) print(X_test.shape) classifier = KNearestNeighbor() classifier.train(X_train, y_train) dists = classifier.compute_distances_two_loops(X_test) print(dists.shape) plt.imshow(dists, interpolation='none') plt.show() # Now implement the function predict_labels and run the code below: # k=1时 y_test_pred = classifier.predict_labels(dists, k=1) # Compute and print the fraction of correctly predicted examples num_correct = np.sum(y_test_pred == y_test) accuracy = float(num_correct) / num_test
import csv
from distance import manhattan_distance, geometric_distance
from datapoint import DataPoint
from k_nearest_neighbor import KNearestNeighbor
from math import *
def load_dataset(filename):
print("loading file " + filename + "...")
with open(filename, 'rb') as csv_file:
return [DataPoint(item[1:], item[0]) for item in csv.reader(csv_file, delimiter=',')]
training_set = load_dataset('train.csv')
test_set = load_dataset('test.csv')
for test_item in test_set:
knn = KNearestNeighbor(test_item, manhattan_distance, 3)
for training_item in training_set:
knn.check(training_item)
correct = test_item.value == closest_neighbor.value
if not correct:
test_item.show()
closest_neighbor.show()
cifar_10_dir = './cifar-10-batches-py'
x_train, y_train, x_test, y_test = load_cifar10(cifar_10_dir)
print('train_data_shape:', x_train.shape)
print('train_labels_shape:', y_train.shape)
print('test_data_shape:', x_test.shape)
print('test_labels_shape:', y_test.shape)
x_train = x_train.reshape(x_train.shape[0], -1)
x_test = x_test.reshape(x_test.shape[0], -1)
num_train = x_train.shape[0]
num_test = x_test.shape[0]
# num_train = 5000
# mask = range(num_train)
# x_train = x_train[mask]
# y_train = y_train[mask]
# num_test = 500
# mask = range(num_test)
# x_test = x_test[mask]
# y_test = y_test[mask]
classifier = KNearestNeighbor()
classifier.train(x_train, y_train)
dicts = classifier.compute_distance(x_test)
y_test_pred = classifier.predict_labels(dicts, k=10)
num_correct = np.sum(y_test_pred == y_test)
accuracy = num_correct / num_test
print('got %d / %d correct => accuracy: %f' %
(num_correct, num_test, accuracy))
def run_model_with_cross_validation(model_name, knn_mode, k_number):
#GET DATA
#- expect data_0 ... data_4
data_groups = list()
data_groups.append(np.loadtxt('data_0', delimiter=','))
data_groups.append(np.loadtxt('data_1', delimiter=','))
data_groups.append(np.loadtxt('data_2', delimiter=','))
data_groups.append(np.loadtxt('data_3', delimiter=','))
data_groups.append(np.loadtxt('data_4', delimiter=','))
NUM_GROUPS = len(data_groups)
#For each data_group, train on all others and test on me
culminating_result = 0;
for test_group_id in range(NUM_GROUPS):
#Form training data as 4/5 data
train_data = np.array([])
for train_group_id in range(len(data_groups)):
if (train_group_id != test_group_id):
#Initialize train_data if necessary
if (train_data.size == 0):
train_data = np.copy(data_groups[train_group_id])
else:
train_data = np.concatenate(
(train_data, data_groups[train_group_id]), axis=0)
print('train_data, group ', str(test_group_id), 'length: ', len(train_data))
print(train_data)
test_data = data_groups[test_group_id]
result = 0
model = None
if (model_name == 'knn'):
model = KNearestNeighbor(train_data, k_number)
model.train(train_data)
print('KNN train data length', len(model.data))
result = model.test(test_data, knn_mode)
elif (model_name == 'c_knn'):
model = CondensedKNearestNeighbor(train_data, k_number)
#Mode is always majority...this is not used for regression
mode = "majority"
model.train(train_data)
print('condensed KNN train data length', len(model.data))
result = model.test(test_data, mode)
else:
print('error - ', model_name, ' is not a supported model')
return
print('test_data, group ', str(test_group_id), 'length:', len(test_data))
print(test_data)
print()
print('result of iteration ' + str(test_group_id))
print(result)
print()
culminating_result = culminating_result + result
final_average_result = culminating_result / NUM_GROUPS
print()
print('final average result:')
print(final_average_result)
print()
return final_average_result
data_file = open(test[0])
print("Running with %s" % test[0])
for line in data_file:
line_split = line.split(',')
data_instances.append(map(float, line_split))
data_instances = np.array(data_instances)
np.random.shuffle(data_instances)
# 5 fold cross validation
learner_type = "CLASSIFICATION"
fold_size = data_instances.shape[0] / 5
data_indices = [idx for idx in range(data_instances.shape[0])]
for k in range(1, 100, 5):
total_performance = 0.0
for holdout_fold_idx in range(5):
kNN_model = KNearestNeighbor(k, learner_type)
kNN_model.train(data_instances[ \
np.array( \
np.setdiff1d(data_indices, data_indices[ \
fold_size * holdout_fold_idx : \
fold_size * holdout_fold_idx + fold_size]))])
kNN_model.condense_training_data()
# predict test data using k-NN and average performance
predictions = kNN_model.predict( \
data_instances[ \
fold_size * holdout_fold_idx : \
fold_size * holdout_fold_idx + fold_size])
successes = fold_size - \
sum(abs(
predictions - \
data_instances[
def __init__(self, data, k_number): KNearestNeighbor.__init__(self, data, k_number)
# 可视化图像
VisualizeImage(X_train, y_train)
input('Enter any key to Cross-validation...')
# 创建用于超参数调优的交叉验证集（也可以验证集，因为数据量还是很大的）
num_training = 5000
X_tr = X_train[:num_training, ::]
X_tr = np.reshape(X_tr, (X_tr.shape[0], -1))
y_tr = y_train[:num_training]
# print(X_tr.shape, y_tr.shape)
num_testing = 500
X_te = X_test[:num_testing, ::]
X_te = np.reshape(X_te, (X_te.shape[0], -1))
y_te = y_test[:num_testing]
# print(X_te.shape, y_te.shape)
# 交叉验证确定参数K
Cross_validation(X_tr, y_tr)
input('Enter any key to train model...')
# 训练完整数据集(这里就以5000个数据集作为完整训练集，500个数据集作为测试集(60000个数据电脑内存吃不消), k值根据图显示10应为最佳)
classify = KNearestNeighbor()
classify.train(X_tr, y_tr)
y_te_pred = classify.predict(X_te, k=10)
accuracy = np.sum(y_te_pred == y_te) / float(X_te.shape[0])
print('最终测试： '
' K = %d, accuracy = %.3f' % (10, accuracy))
fold_size * holdout_fold_idx + fold_size]))])
centroids = kmeans_model.get_centroids()
for cluster_idx in xrange(len(clusters)):
ave_label = 0.0
for instance in clusters[cluster_idx]:
ave_label += instance[-1]
if len(clusters[cluster_idx]) > 0:
ave_label = ave_label / len(clusters[cluster_idx])
if learner_type == "CLASSIFICATION":
ave_label = int(round(ave_label))
centroids[cluster_idx].append(ave_label)
# for classification, vote to determine centroid classification
# for regression, average to find centroid estimate
# feed centroids into k-NN as training data
kNN_model = KNearestNeighbor(best_ks[test[0]], learner_type)
kNN_model.train(centroids)
# predict test data using k-NN and average performance
predictions = kNN_model.predict( \
data_instances[ \
fold_size * holdout_fold_idx : \
fold_size * holdout_fold_idx + fold_size])
if kNN_model.learner_type == "CLASSIFICATION":
successes = fold_size - \
sum(abs(
predictions - \
data_instances[
fold_size * holdout_fold_idx :
fold_size * holdout_fold_idx + fold_size,-1]))
performance = successes / fold_size
elif kNN_model.learner_type == "REGRESSION":
file_name = train_files[i]
file_str = file_name.split('.')[0]
class_num = int(file_str.split('_')[0])
digits_labels.append(class_num)
training_mat[i, :] = img2vector('../../data/digits/trainingDigits/%s' %
file_name)
# 使用训练集进行训练
classifier.fit(training_mat, digits_labels)
test_files = os.listdir('../../data/digits/testDigits')
err_cnt = 0
m_test = len(test_files)
for i in range(m_test):
file_name = test_files[i]
file_str = file_name.split('.')[0]
class_num = int(file_str.split('_')[0])
vec_under_test = img2vector('../../data/digits/testDigits/%s' %
file_name)
classifier_result = classifier.predict(vec_under_test)
print('the classifier came back with: %d, the real answer is: %d' %
(classifier_result, class_num))
if classifier_result != class_num:
err_cnt += 1
print('the total number of errors is: %d' % err_cnt)
print('the total error rate is: %f' % (err_cnt / float(m_test)))
if __name__ == '__main__':
knn = KNearestNeighbor(k=10)
digits_classify(classifier=knn)
# subsample train data. array indexing
X_train = X_train[mask]
y_train = y_train[mask]
num_test = 500
mask = range(num_test)
# subsample test data. array indexing
X_test = X_test[mask]
y_test = y_test[mask]
# reshape data to rows
X_train = np.reshape(X_train, (X_train.shape[0], -1))
X_test = np.reshape(X_test, (X_test.shape[0], -1))
# initial classifer object
knn_classifer = KNearestNeighbor()
# train with training set
# knn_classifer.train(X_train, y_train)
"""
# compute distances for test set
"""
# dists = knn_classifer.compute_distances_two_loops(X_test)
# dists_one_loop = knn_classifer.compute_distances_one_loop(X_test)
# dists_no_loop = knn_classifer.compute_distances_no_loops(X_test)
"""
# test one loop computation
"""
# difference = np.linalg.norm(dists - dists_one, ord='fro')
# print('One loop Difference was: %f' % (difference, ))
"""
test no loop computation
#Parameters: #a : array_like #Array to be reshaped. #newshape : int or tuple of ints #The new shape should be compatible with the original shape. If an integer, then the result will be a 1-D array of that length. One shape dimension can be -1. from k_nearest_neighbor import KNearestNeighbor # Create a kNN classifier instance. # Remember that training a kNN classifier is a noop: # the Classifier simply remembers the data and does no further processing classifier = KNearestNeighbor() classifier.train(X_train, y_train) #Open cs231n/classifiers/k_nearest_neighbor.py and implement # compute_distances_two_loops. # Test your implementation: dists = classifier.compute_distances_two_loops(X_test) print(dists.shape) # We can visualize the distance matrix: each row is a single test example and # its distances to training examples plt.imshow(dists,interpolation='none') plt.show() # Now implement the function predict_labels and run the code below:
mask=list(range(num_training))
X_train=X_train[mask]
y_train=y_train[mask]
num_test=500
mask=list(range(num_test))
X_test=X_test[mask]
y_test=y_test[mask]
#Reshape the image data into rows
X_train=np.reshape(X_train,(X_train.shape[0],-1))
X_test=np.reshape(X_test,(X_test.shape[0],-1))
print(X_train.shape,X_test.shape)
#Create a KNN classifier instance,k=1
classifier=KNearestNeighbor()
classifier.train(X_train,y_train)
"""
dists=classifier.compute_distance_two_loops(X_test)
print('dists.shape is')
print(dists.shape)
#plt.imshow(dists,interpolation='none')
#plt.savefig('/home/hongyin/file/cs231n-assignment1/picFaster.jpg')
y_test_pred=classifier.predict_labels(dists,k=1)
num_correct=np.sum(y_test_pred==y_test)
accuracy=float(num_correct)/num_test
print('Got %d / %d correct => accuracy: %f' % (num_correct, num_test, accuracy))
#k=5
y_test_pred = classifier.predict_labels(dists, k=5)
import sklearn
if __name__ == '__main__':
train_path = "/Users/zxj/Desktop/Mini1/train.pkl"
train_data = pickle.load(open(train_path, "rb"))
# Fixed_parameters
# Please do not change the fixed parameters
val_ratio = 0.2
# student_parameters
# You may want to change these in your experiment later.
train_ratio = 1.0 # we split the train_data into 0.8:training
train_num = int(train_data['data'].shape[0] * train_ratio *
(1.0 - val_ratio))
val_num = -1 * int(train_data['data'].shape[0] * train_ratio * val_ratio)
KNN_classifier = KNearestNeighbor()
KNN_classifier.train(train_data['data'][:train_num],
train_data['target'][:train_num])
dists = KNN_classifier.compute_distances(train_data['data'][val_num:, :])
k_choices = [2, 3, 5, 7, 9, 11, 15, 19]
for k in k_choices:
y_test_pred = KNN_classifier.predict_labels(dists, k)
num_correct = np.sum(y_test_pred == train_data['target'][val_num:])
accuracy = float(num_correct) / (-1 * val_num)
print(
'For K= %d and train_ratio= %f, Got %d / %d correct => VAL_accuracy: %f'
% (k, train_ratio, num_correct, -1 * val_num, accuracy))
num_test = 500
X_test = X_test[:num_test]
Y_test = Y_test[:num_test]
num_folds = 5
k_choices = [1, 3, 5, 8, 10, 12, 15, 20, 50, 100]
x_train_folds = []
y_train_folds = []
x_train_folds = np.array_split(X_train, num_folds)
y_train_folds = np.array_split(Y_train, num_folds)
k_to_accuracies = {}
classifier = KNearestNeighbor()
for k in k_choices:
accuracies = np.zeros(num_folds)
for fold in range(num_folds):
temp_X = x_train_folds[:]
temp_y = y_train_folds[:]
x_validate_fold = temp_X.pop(fold)
y_validate_fold = temp_y.pop(fold)
temp_X = np.array([y for x in temp_X for y in x])
temp_y = np.array([y for x in temp_y for y in x])
classifier.train(temp_X, temp_y)
y_test_pred = classifier.predict(x_validate_fold, k=k)
num_correct = np.sum(y_test_pred == y_validate_fold)
accuracy = float(num_correct) / num_test