def getBestK(X_train, y_train, X_val, y_val, nns=[30], print_train=True, print_val=True):
acc_train = np.zeros((1, len(nns)))
acc_val = np.zeros((1, len(nns)))
for j in range(0, len(nns)):
print j
sys.stdout.flush()
knn = KNNClassifier(nns[j])
knn.train(X_train, y_train)
# acc_train[0, j] = np.mean(knn.predict(X_train) == y_train)
print acc_train[0, j]
sys.stdout.flush()
y_pred = knn.predict(X_val)
acc_val[0, j] = np.mean(y_pred == y_val)
print acc_val[0, j]
sys.stdout.flush()
print "Confusion matrix:"
print confusion_matrix(y_pred, y_val)
if print_train:
print (acc_train)
if print_val:
print (acc_val)
best_val = np.max(acc_val)
best_rate, best_reg = np.where(acc_val == np.amax(acc_val))
return (best_rate[0], best_reg[0]), knn
def knn():
X, y = make_blobs(centers=4, n_samples=500, n_features=2,
shuffle=True)
model = KNNClassifier(K=4)
model.fit(X, y)
res = model.predict(X)
print(np.mean(res == y))
def main():
curr_dir = os.path.dirname(__file__)
csv_file = os.path.join(curr_dir, 'data/play.csv')
test = pd.Series({
'Tempo': 'Chuva',
'Temperatura': 'Quente',
'Humidade': 'Normal',
'Vento': 'Forte'
})
df = pd.read_csv(csv_file, index_col='Dia')
X, y = df.loc[:, df.columns != 'Jogar'], df['Jogar']
clf = KNNClassifier(k=1)
clf.fit(X, y)
print(f'RESULT k = {clf.k} ::',
'Jogar' if clf.predict(test) else 'Não Jogar')
clf.k = 3
print(f'RESULT k = {clf.k} ::',
'Jogar' if clf.predict(test) else 'Não Jogar')
print()
print('DISTANCES')
print(clf._gen_distances(test))
def main():
df = pd.read_csv('./diabetes.csv')
normalized_data = normalize(df, 'Outcome')
knn_classifier = KNNClassifier()
for k in [5, 10]:
results = {}
for nn in [3, 5, 7]:
knn_classifier.nn = nn
knn_classifier.cross_validate(
normalized_data,
k, # k folds
1 # r
)
results[str(nn)] = [
knn_classifier.accuracy, knn_classifier.f1_score
]
plot_results(results)
for nn in [3, 5, 7]:
knn_classifier.nn = nn
knn_classifier.cross_validate(
normalized_data,
10, # k folds
10 # r
)
print('\nGlobal accuracy: %.3f (%.3f)' %
(knn_classifier.accuracy, knn_classifier.accuracy_std))
print('Global f1 score accuracy: %.3f (%.3f)\n' %
(knn_classifier.f1_score, knn_classifier.f1_score_std))
results[str(nn)] = [knn_classifier.accuracy, knn_classifier.f1_score]
plot_results(results)
def test_knn(self):
knn_model = KNNClassifier(all_monomials_with_maximum_degrees([1, 1, 1]), 1)
knn_model.train(np.array([[1], [10], [2], [30]]), np.array([1, 0, 1, 0]))
self.assertEqual(
knn_model._trained,
True,
"Initially False. Change this property to True after self.train() is called",
)
self.assertIn(
type(knn_model.predict(np.array([1]))),
(np.float64, float),
"Return type of predict() is np.float64 or float",
)
self.assertIn(
type(
knn_model.evaluate(
np.array([[1], [10], [2], [20]]), np.array([1, 0, 1, 0])
)
),
(float, np.float64, int, np.int64),
"Return type of evaluate() is np.float64 or float",
)
self.assertIn(
knn_model.evaluate(
np.array([[1], [10], [2], [30]]), np.array([1, 0, 1, 0])
),
(1.0,1),
"evaluate() for same data returns 1.0",
)
def test():
train_data, train_labels = mnist.load_mnist(mode='train', path='data/')
# test_data, test_labels = mnist.load_mnist(mode='test', path='data/')
errors = np.array(
knn.tune_hyperparams(train_data[:1000], train_labels[:1000]))
X = errors[:, 0]
Y = errors[:, 1]
Z = errors[:, 2]
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.scatter(X, Y, Z, c='r', marker='o')
ax.set_xlabel('X Label')
ax.set_ylabel('Y Label')
ax.set_zlabel('Z Label')
plt.show()
def run(k,
mode='knn',
distance='euclidean',
keep_punc=False,
keep_stopwords=False):
from knn import KNNClassifier
from ncc import NearestCentroidClassifier
if mode == 'knn':
clf = KNNClassifier(k,
distance=distance,
keep_stopwords=keep_stopwords,
keep_punc=keep_punc)
clf.score()
elif mode == 'ncc':
clf = NearestCentroidClassifier(k,
distance=distance,
keep_stopwords=keep_stopwords,
keep_punc=keep_punc)
clf.score()
def training(q, DTrain, RTrain, **args):
'''
@input topic document q ∈ Q, training collection Dtrain, judgments Rtrain, and
optional arguments on the classification process
@behavior learns a classification model to predict the relevance of documents on the
topic q using Dtrain and Rtrain, where the training process is subjected to
proper preprocessing, classifier’s selection and hyperparameterization
@output q-conditional classification model
'''
classifier_type = args.get('classifier_type')
if classifier_type == 'logistic':
hyper_parameters = args.get('hyper_paremeters')
classifier = LogisticClassifier(hyper_parameters=hyper_parameters)
elif classifier_type == 'XGBOOST':
classifier = XGBOOSTClassifier()
elif classifier_type == 'MLP':
classifier = MLPerceptronClassifier()
elif classifier_type == 'KNN':
classifier = KNNClassifier()
classifier.train(q, DTrain, RTrain)
return classifier
print(_, nFeats)
# Values of parameter k to iterate over.
K_VALS = [3, 5, 7, 9, 11, 13, 15]
starttime = time.time()
# Repeat each trial 10 times.
for i in range(0, 10):
x_train, x_test, y_train, y_test = train_test_split(X, y,\
test_size=0.2)
"""
Try non-optimized methods.
"""
# Vanilla KNN.
for k in K_VALS:
reg = KNNClassifier(x_train, y_train, k)
y_pred = reg.predict(x_test)
mse_iter = accuracy(y_test, y_pred)
print("xx,knn,", k, ",", mse_iter)
# Distance-weighted KNN.
for k in K_VALS:
reg = DwKNNClassifier(x_train, y_train, k)
y_pred = reg.predict(x_test)
mse_iter = accuracy(y_test, y_pred)
print("xx,dknn,", k, ",", mse_iter)
"""
PCA with KNN.
"""
pca = PCA(n_components=4)
pca.fit(x_train.copy())
# Perform dimensionality reduction #
X_train_reduced = svd.fit_transform(X_train_tfidf)
# Keep results #
prec, rec, f1, accu = ([] for i in range(4))
# Use 10-fold and find metrics #
for train, test in kf.split(X_train_reduced, X_train_le):
X_train = X_train_reduced[train]
y_train = X_train_le[train]
X_test = X_train_reduced[test]
y_test = X_train_le[test]
clf_KNN = KNNClassifier(100)
# Train model #
clf_KNN.fit(X_train, y_train)
# Predict categories #
y_pred = clf_KNN.predict(X_test)
# Save scores #
prec.append(precision_score(y_test, y_pred, average='macro'))
rec.append(recall_score(y_test, y_pred, average='macro'))
f1.append(f1_score(y_test, y_pred, average='macro'))
accu.append(accuracy_score(y_test, y_pred))
# Print results to csv #
Evaluation_metric_df = pd.read_csv('EvaluationMetric_10fold.csv', sep="\t")
#if (nFeats > 15) or (_ > 4000):
# continue
print("Number of samples: ",_, "Number of features: ", nFeats)
#print("X :")
#print(X)
#print("y :")
#print(y)
print("Splitting training and test sets:")
#Test without scaling
print("Testing without scaling")
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
x_train, x_verif, y_train, y_verif = train_test_split(x_train, y_train, test_size=0.33)
clf = KNNClassifier(x_train, y_train, 5)
y_pred = clf.predict(x_test)
print("Accuracy = ", accuracy_score(y_test, y_pred))
#Run GA to find best weights
N_init_pop = 30
_, nFeats = np.shape(x_train)
weight_pso = GBestPSO(nFeats, N_init_pop)
pos = weight_pso.get_positions()
pbest = weight_pso.get_pbest()
pbest_metric_array = np.empty(N_init_pop)
pos_metric_array = np.empty(N_init_pop)
#Set pbest metrics
for i in range(len(pbest)):
import numpy as np import matplotlib.pyplot as plt from sklearn import datasets from knn import KNNClassifier from module_selection import train_test_split iris = datasets.load_iris() #print(iris.keys()) X = iris.data y = iris.target X_train, X_test, y_train, y_test = train_test_split(X, y) my_knn_clf = KNNClassifier(k=3) my_knn_clf.fit(X_train, y_train) y_predict = my_knn_clf.predict(X_test) #accuracy accuracy = sum(y_predict == y_test) / len(y_test) print(accuracy) # using sklean KNeighborsClassifier and model_selection """ import numpy as np import matplotlib.pyplot as plt from sklearn import datasets from sklearn.neighbors import KNeighborsClassifier from sklearn.model_selection import train_test_split iris = datasets.load_iris()
def lbest_pso_run_ent(x_train, y_train, x_test, y_test, x_verif, y_verif, k):
#Run PSO to find best weights
N_init_pop = 50
_, nFeats = np.shape(x_train)
weight_pso = LBestPSO(nFeats, N_init_pop)
pos = weight_pso.get_positions()
pbest = weight_pso.get_pbest()
pbest_metric_array = np.empty(N_init_pop)
pos_metric_array = np.empty(N_init_pop)
#Set pbest metrics
for i in range(len(pbest)):
#Scale input data
scaled_x_train = np.multiply(x_train, pbest[i])
#Scale verificaion data
scaled_x_verif = np.multiply(x_verif, pbest[i])
#Method 1
clf = KNNClassifier(scaled_x_train, y_train, k)
neighbors = clf.find_all_neighbors(scaled_x_verif)
nbh_ent = clf.find_neighborhood_entropy(neighbors)
pbest_metric_array[i] = nbh_ent
weight_pso.set_pbest_fitness(pbest_metric_array)
#Set pos metrics
for i in range(len(pbest)):
#Scale input data
scaled_x_train = np.multiply(x_train, pos[i])
#Scale verificaion data
scaled_x_verif = np.multiply(x_verif, pos[i])
#Method 1
clf = KNNClassifier(scaled_x_train, y_train, k)
neighbors = clf.find_all_neighbors(scaled_x_verif)
nbh_ent = clf.find_neighborhood_entropy(neighbors)
pos_metric_array[i] = nbh_ent
weight_pso.set_p_fitness(pos_metric_array)
#Set initial gbest.
weight_pso.set_init_best(pos_metric_array)
count = 0
while (count < 50):
count += 1
weight_pso.optimize()
#get_population
weight_pop = weight_pso.get_positions()
metric_array = np.empty(N_init_pop)
#evaluate and set fitness
for i in range(len(weight_pop)):
#Scale input data
scaled_x_train = np.multiply(x_train, weight_pop[i])
#Scale verificaion data
scaled_x_verif = np.multiply(x_verif, weight_pop[i])
#Method 1
clf = KNNClassifier(scaled_x_train, y_train, k)
neighbors = clf.find_all_neighbors(scaled_x_verif)
nbh_ent = clf.find_neighborhood_entropy(neighbors)
metric_array[i] = nbh_ent
weight_pso.set_p_fitness(metric_array)
weight_pso.set_best(metric_array)
#get_best_sol
best_metric = weight_pso.get_gbest_fit()
best_weights = weight_pso.get_gbest()
# Concatenate training and verification sets.
x_train = np.concatenate((x_train, x_verif), axis = 0)
y_train = np.concatenate([y_train, y_verif])
# Print the results of KNN.
clf = KNNClassifier(np.multiply(x_train, best_weights), y_train, k)
y_pred = clf.predict(np.multiply(x_test, best_weights))
mse_iter = accuracy(y_test, y_pred)
print("lbest-pso-ent,knn,", k, ",", mse_iter)
# Print the results of KNN.
clf = DwKNNClassifier(np.multiply(x_train, best_weights), y_train, k)
y_pred = clf.predict(np.multiply(x_test, best_weights))
mse_iter = accuracy(y_test, y_pred)
print("lbest-pso-ent,dknn,", k, ",", mse_iter)
points.append([float(inp[0]), float(inp[1]), int(inp[2])])
random.shuffle(points)
n_points = len(points)
points = chunkify(points)
acc1, acc2 = [], []
for i in range(NUMBER_OF_FOLDS):
train, test = [], []
for j in range(NUMBER_OF_FOLDS):
if (i != j):
train.extend(points[j])
else:
test = points[j]
knn = KNNClassifier()
svm = SVM()
train2 = copy.deepcopy(train)
knn.fit(train)
svm.fit_transform(train2)
k_right, s_right = 0, 0
for point in test:
k_pred = knn.predict([point[0], point[1]])
s_pred = svm.predict([point[0], point[1]])
s_pred = (s_pred + 1) / 2
if (k_pred == point[2]):
k_right = k_right + 1
if (s_pred == point[2]):
s_right = s_right + 1
acc1.append(float(k_right / len(test)))
acc2.append(float(s_right / len(test)))
train_imgs = utils.read_folder(TRAIN_DIR, 0, ntrain, flatten=False)
print ("\nDone!")
sys.stdout.flush()
X = train_imgs
X = X.reshape((ntrain, -1))
#X = np.insert(X, 0, 1.0, axis = 1)
y = utils.read_labels('trainLabels.csv', 0, ntrain)
X_train, X_val, y_train, y_val = cross_validation.train_test_split(X, y, test_size = 0.1)
nns = [1]
#utils.getBestK(X_train, y_train, X_val, y_val, nns)
knn = KNNClassifier(nns[0])
knn.train(X_train, y_train)
print "X_val shape: ", X_val.shape
print "y_val shape: ", y_val.shape
pred = knn.predict(X_val)
print "Accuracy: ", np.mean(pred == y_val)
#uncomment this to visualize knn prediction - 10 examples from each class
"""
examples = np.zeros((10,10,32,32,3))
for i in range(10):
examples[i] = ((X_val[pred==i])[0:10]).reshape(10,32,32,3)
num_classes = len(classes)
nexamples = 10
for y, cls in enumerate(classes):
idxs = np.arange(nexamples)
for i, idx in enumerate(idxs):
else:
matrix = evaluator.compute().cpu()
plotconfmat(dataset_name.split(","), matrix,
rootname + ".png")
torch.save(matrix, rootname + ".pt")
# T.to_pil_image(matrix).save(f"matrix_{classifier_name}_dataset{i:02}.png")
import json
with open(rootname + ".json", "w+") as f:
json.dump(outputs, f)
solver.cpu()
classifiers = {
"knn5": KNNClassifier(k=5),
"knn10": KNNClassifier(k=10),
# "relpnet_1": M.RelationNetClassifier_Protonet1(simnet_channels=[128, 64, 32]),
# "relation_1": M.RelationNetClassifier(in_channels=416, feature_channels=[10], simnet_channels=[32, 16, 4]),
# "reg_relation_1": M.RelationNetClassifier(in_channels=416, feature_channels=[10], simnet_channels=[32, 16, 4]),
# "protonet_1": M.ProtonetClassifier(in_channels=416, mid_channels=[], out_channels=32),
# "protonet_2": M.ProtonetClassifier(in_channels=416, mid_channels=[64], out_channels=32),
# "protonet_3": M.ProtonetClassifier(in_channels=416, mid_channels=[128, 64], out_channels=32),
# "protonet_4": M.ProtonetClassifier(in_channels=416, mid_channels=[256, 128, 64], out_channels=32),
# "protonet_bottleneck_end": M.ProtonetClassifier(in_channels=416, mid_channels=[256, 128, 64], out_channels=10),
# "protonet_bottleneck_mid": M.ProtonetClassifier(in_channels=416, mid_channels=[128, 32, 128], out_channels=32),
# "simnet_simple": M.SimnetClassifier(in_channels=416, channels=[10]),
def ga_run_std(x_train, y_train, x_test, y_test, x_verif, y_verif, k):
# Run GA to find best weights.
N_init_pop = 50
N_crossover = 50
N_selection = 20
improv_thresh = 1e-3
_, nFeats = np.shape(x_train)
weight_ga = GeneticAlgorithm(nFeats, N_init_pop, mu=0.1)
weight_pop = weight_ga.get_population()
metric_array = np.empty(N_init_pop)
# Create the initial population.
for i in range(len(weight_pop)):
# Scale input data
scaled_x_train = np.multiply(x_train, weight_pop[i])
# Scale verificaion data
scaled_x_verif = np.multiply(x_verif, weight_pop[i])
# Classifier.
clf = KNNClassifier(scaled_x_train, y_train, k)
neighbors = clf.find_all_neighbors(scaled_x_verif)
nbh_ent = clf.find_neighborhood_std(neighbors)
metric_array[i] = nbh_ent
# Update fitness in GA object.
weight_ga.set_fitness(metric_array)
weight_ga.selection(N_selection)
new_best_metric = 2.5
# while (best_metric - new_best_metric) > improv_thresh:
count = 0
while (count < 20):
count += 1
best_metric = new_best_metric
# Crossover.
weight_ga.crossover(N_crossover)
# Get new population.
weight_pop = weight_ga.get_population()
metric_array = np.empty(N_crossover)
# Evaluate and set fitness.
for i in range(len(weight_pop)):
# Scale input data
scaled_x_train = np.multiply(x_train, weight_pop[i])
# Scale verificaion data
scaled_x_verif = np.multiply(x_verif, weight_pop[i])
# Classifier.
clf = KNNClassifier(scaled_x_train, y_train, k)
neighbors = clf.find_all_neighbors(scaled_x_verif)
nbh_ent = clf.find_neighborhood_std(neighbors)
metric_array[i] = nbh_ent
# Update fitness in GA object
weight_ga.set_fitness(metric_array)
# get_best_sol
best_weights, new_best_metric = weight_ga.best_sol()
#print("Metric of this iteration are: ", new_best_metric)
weight_ga.selection(N_selection)
# print("Best weights = ", best_weights, "\tBest metric = ", new_best_metric)
# Test with scaling after GA
# Concatenate training and verification sets.
x_train = np.concatenate((x_train, x_verif), axis=0)
y_train = np.concatenate([y_train, y_verif])
# Print the results of KNN.
clf = KNNClassifier(np.multiply(x_train, best_weights), y_train, k)
y_pred = clf.predict(np.multiply(x_test, best_weights))
acc = accuracy(y_test, y_pred)
print("ga-std,knn,", k, ",", acc)
# Print the results of KNN.
clf = DwKNNClassifier(np.multiply(x_train, best_weights), y_train, k)
y_pred = clf.predict(np.multiply(x_test, best_weights))
mse_iter = accuracy(y_test, y_pred)
print("ga-std,dknn,", k, ",", mse_iter)
