class Number_recognition:
def __init__(self, test_images, test_labels, k):
self.test_images = test_images;
self.test_labels = test_labels;
self.k = k;
def load(self, path_img, path_lbl):
with open(path_lbl, 'rb') as file:
magic, size = struct.unpack(">II", file.read(8))
if magic != 2049:
raise ValueError('Magic number mismatch, expected 2049,'
'got %d' % magic)
labels = array("B", file.read())
labels_result = np.zeros(shape=(size))
for i in xrange(size):
labels_result[i] = labels[i]
with open(path_img, 'rb') as file:
magic, size, rows, cols = struct.unpack(">IIII", file.read(16))
print "rows: " + str(rows) + " cols: " + str(cols)
if magic != 2051:
raise ValueError('Magic number mismatch, expected 2051,'
'got %d' % magic)
image_data = array("B", file.read())
images = np.zeros(shape=(size,rows*cols))
for i in xrange(size):
images[i][:] = image_data[i*rows*cols : (i+1)*rows*cols]
return images, labels_result
def load_train(self):
ims, labels = self.load( self.test_images, self.test_labels)
self.test_images = ims
self.test_labels = labels
labels_numbers = MulticlassLabels(self.test_labels)
feats = RealFeatures(self.test_images.T)
dist = EuclideanDistance()
self.knn = KNN(self.k, dist, labels_numbers)
self.knn.train(feats)
def predict(self, image):
feats_test = RealFeatures(image. T)
pred = self.knn.apply_multiclass(feats_test)
return pred[:]
def run_knn(Xtrain,Ytrain,Xtest,Ytest):
prod_features = RealFeatures(Xtrain)
prod_labels = MulticlassLabels(Ytrain)
test_features = RealFeatures(Xtest)
test_labels = MulticlassLabels(Ytest)
if os.path.exists(".lmnn_model30000_5_reg05_cor20"):
print "Using LMNN distance"
lmnn = LMNN()
sf = SerializableAsciiFile(".lmnn_model30000_5_reg05_cor20", 'r')
lmnn.load_serializable(sf)
diagonal = np.diag(lmnn.get_linear_transform())
#print('%d out of %d elements are non-zero.' % (np.sum(diagonal != 0), diagonal.size))
#diagonal = lmnn.get_linear_transform()
np.set_printoptions(precision=1,threshold=1e10,linewidth=500)
#lmnn.set_diagonal(True)
dist = lmnn.get_distance()
else:
dist = EuclideanDistance()
# classifier
knn = KNN(K, dist, prod_labels)
#knn.set_use_covertree(True)
parallel = knn.get_global_parallel()
parallel.set_num_threads(4)
knn.set_global_parallel(parallel)
knn.train(prod_features)
print "Classifying test set..."
pred = knn.apply_multiclass(test_features)
print "Accuracy = %2.2f%%" % (100*np.mean(pred == Ytest))
cm = build_confusion_matrix(Ytest, pred, NCLASSES)
#save_confusion_matrix(cm)
#cm = load_confusion_matrix()
print "Confusion matrix: "
print cm
#plot_confusion_matrix(cm)
#results = predict_class_prob(pred, cm)
#nn = build_neighbours_matrix(knn, prod_labels)
#results = predict_class_from_neighbours(nn)
#print "Log loss: " + str(calculate_log_loss(results, Ytest))
#print_prediction_output(results)
return cm
# load LMNN
if os.path.exists(".lmnn_model30000_5_reg05_cor20"):
sf = SerializableAsciiFile(".lmnn_model30000_5_reg05_cor20", 'r')
lmnn = LMNN()
lmnn.load_serializable(sf)
diagonal = np.diag(lmnn.get_linear_transform())
print('%d out of %d elements are non-zero.' % (np.sum(diagonal != 0), diagonal.size))
#print diagonal
dist = lmnn.get_distance()
else:
dist = EuclideanDistance()
cm = load_confusion_matrix()
print cm
# classifier
knn = KNN(k, dist, prod_labels)
parallel = knn.get_global_parallel()
parallel.set_num_threads(4)
knn.set_global_parallel(parallel)
knn.train(prod_features)
print "Classifying test set..."
pred = knn.apply_multiclass(test_features)
results = predict_class_prob(pred, cm)
print_prediction_output(results)
def evaluate(labels,
feats,
params={
'n_neighbors': 2,
'use_cover_tree': 'True',
'dist': 'Manhattan'
},
Nsplit=2):
"""
Run Cross-validation to evaluate the KNN.
Parameters
----------
labels: 2d array
Data set labels.
feats: array
Data set feats.
params: dictionary
Search scope parameters.
Nsplit: int, default = 2
The n for n-fold cross validation.
all_ks: range of int, default = range(1, 21)
Numbers of neighbors.
"""
k = params.get('n_neighbors')
use_cover_tree = params.get('use_cover_tree') == 'True'
if params.get('dist' == 'Euclidean'):
func_dist = EuclideanDistance
else:
func_dist = ManhattanMetric
split = CrossValidationSplitting(labels, Nsplit)
split.build_subsets()
accuracy = np.zeros(Nsplit)
acc_train = np.zeros(accuracy.shape)
time_test = np.zeros(accuracy.shape)
for i in range(Nsplit):
idx_train = split.generate_subset_inverse(i)
idx_test = split.generate_subset_indices(i)
feats.add_subset(idx_train)
labels.add_subset(idx_train)
dist = func_dist(feats, feats)
knn = KNN(k, dist, labels)
knn.set_store_model_features(True)
if use_cover_tree:
knn.set_knn_solver_type(KNN_COVER_TREE)
else:
knn.set_knn_solver_type(KNN_BRUTE)
knn.train()
evaluator = MulticlassAccuracy()
pred = knn.apply_multiclass()
acc_train[i] = evaluator.evaluate(pred, labels)
feats.remove_subset()
labels.remove_subset()
feats.add_subset(idx_test)
labels.add_subset(idx_test)
t_start = time.clock()
pred = knn.apply_multiclass(feats)
time_test[i] = (time.clock() - t_start) / labels.get_num_labels()
accuracy[i] = evaluator.evaluate(pred, labels)
feats.remove_subset()
labels.remove_subset()
print accuracy.mean()
return accuracy