def main(unused_argv):
ROOT_PATH = "." # Denotes the current working directory
TRAIN_DATA_DIRECTORY = os.path.join(ROOT_PATH, "/root/leaf_image/DATA/training")
TEST_DATA_DIRECTORY = os.path.join(ROOT_PATH, "/root/leaf_image/DATA/testing")
train_data, train_labels = load_data(TRAIN_DATA_DIRECTORY)
eval_data, eval_labels = load_data(TEST_DATA_DIRECTORY)
# Create the Estimator
mnist_classifier = tf.estimator.Estimator(
model_fn=cnn_model_fn, model_dir="./tmp/model")
# Set up logging for predictions
# Log the values in the "Softmax" tensor with label "probabilities"
tensors_to_log = {"probabilities": "softmax_tensor"}
logging_hook = tf.train.LoggingTensorHook(
tensors=tensors_to_log, every_n_iter=50)
# Train the model
train_input_fn = tf.estimator.inputs.numpy_input_fn(
x={"x": train_data},
y=train_labels,
batch_size=100,
num_epochs=None,
shuffle=True)
mnist_classifier.train(
input_fn=train_input_fn,
steps=100,
hooks=[logging_hook])
def serving_input_receiver_fn():
"""Build the serving inputs."""
inputs = {"x": tf.placeholder(shape=[1, DEFAULT_SIZE, DEFAULT_SIZE, 1], dtype=tf.float32)}
return tf.estimator.export.ServingInputReceiver(inputs, inputs)
export_dir = mnist_classifier.export_savedmodel(
export_dir_base="./model_saved/",
serving_input_receiver_fn=serving_input_receiver_fn)
# Evaluate the model and print results
eval_input_fn = tf.estimator.inputs.numpy_input_fn(
x={"x": eval_data},
y=eval_labels,
num_epochs=2,
shuffle=False)
print(eval_input_fn)
eval_results = mnist_classifier.evaluate(input_fn=eval_input_fn)
print(eval_results)
predict_input_fn = tf.estimator.inputs.numpy_input_fn(
x={"x": train_data[0]},
shuffle=False)
prediction_results = mnist_classifier.predict(predict_input_fn)
for i in prediction_results:
print(i)
print(i['classes'])
def _pipeline_w2v_and_recurentneuralnetwork(self, dir_dataset):
categories = os.listdir(dir_dataset)
# X_train
X_train, y_train = [], []
for category in tqdm(categories):
path = os.path.join(dir_dataset, category, 'evc.train.en')
X_tmp, y_tmp = load_data(path)
for i in range(len(X_tmp)):
X_train.append(X_tmp[i])
y_train.append(categories.index(category))
# X_val
X_val, y_val = [], []
for category in tqdm(categories):
path = os.path.join(dir_dataset, category, 'evc.dev.en')
X_tmp, y_tmp = load_data(path)
for i in range(len(X_tmp)):
X_val.append(X_tmp[i])
y_val.append(categories.index(category))
# X_test
X_test, y_test = [], []
for category in tqdm(categories):
path = os.path.join(dir_dataset, category, 'evc.test.en')
X_tmp, y_tmp = load_data(path)
for i in range(len(X_tmp)):
X_test.append(X_tmp[i])
y_test.append(categories.index(category))
# transform text to vector by word to vec pretrain model
W2V = WordEmbedding().Word2Vec()
X_train = self._text2vecs(W2V, X_train)
X_val = self._text2vecs(W2V, X_val)
X_test = self._text2vecs(W2V, X_test)
y_train = to_categorical(y_train)
y_val = to_categorical(y_val)
input_dim, classes = len(X_train[0][0]), len(categories)
model = rnn_text_classification(input_dim, classes).model
model.fit(X_train, y_train, epochs = 2, batch_size = 16, validation_data=(X_val, y_val))
y_hat = model.predict(X_test)
y_hat = np.argmax(y_hat, axis = 1)
print(classification_report(y_test, y_hat))
return categories, W2V, model
def _pipeline_bow_and_neuralnetwork(self, dir_dataset):
categories = os.listdir(dir_dataset)
# X_train
X_train, y_train = [], []
for category in tqdm(categories):
path = os.path.join(dir_dataset, category, 'evc.train.en')
X_tmp, y_tmp = load_data(path)
for i in range(len(X_tmp)):
X_train.append(X_tmp[i])
y_train.append(categories.index(category))
# X_val
X_val, y_val = [], []
for category in tqdm(categories):
path = os.path.join(dir_dataset, category, 'evc.dev.en')
X_tmp, y_tmp = load_data(path)
for i in range(len(X_tmp)):
X_val.append(X_tmp[i])
y_val.append(categories.index(category))
# X_test
X_test, y_test = [], []
for category in tqdm(categories):
path = os.path.join(dir_dataset, category, 'evc.test.en')
X_tmp, y_tmp = load_data(path)
for i in range(len(X_tmp)):
X_test.append(X_tmp[i])
y_test.append(categories.index(category))
# transform text to vector by word of bag
BoW = WordEmbedding().CountVectorizer()
X_train = BoW.fit_transform(X_train)
X_train = X_train.toarray()
X_val = BoW.transform(X_val).toarray()
X_test = BoW.transform(X_test).toarray()
y_train = to_categorical(y_train)
y_val = to_categorical(y_val)
model = Neural_Network(len(X_train[0]), len(categories)).model
model.fit(X_train, y_train, epochs = 5, batch_size = 16, validation_data=(X_val, y_val))
y_hat = model.predict(X_test)
y_hat = np.argmax(y_hat, axis = 1)
print(classification_report(y_test, y_hat))
return categories, BoW, model
def preprocess_dataset(classes_authorized, components, compression_method, patch_size):
X, train_data, test_data = pp.load_data()
train_data = pp.delete_useless_classes(train_data, classes_authorized)
test_data = pp.delete_useless_classes(test_data, classes_authorized)
print("Before Shuffle: ")
pretty_print_count(train_data, test_data)
train_data, test_data = pp.shuffle_train_test(train_data, test_data)
print("After Shuffle: ")
pretty_print_count(train_data, test_data)
if compression_method is not None:
X, pca = pp.dimensionality_reduction(X, numComponents=components, standardize=False,
compression=compression_method)
# CREATE PATCHES, DELETE 0 VALUES
X_train, X_test, y_train, y_test = pp.patch_1dim_split(X, train_data, test_data, patch_size)
y_train = np_utils.to_categorical(y_train, num_classes=9)
y_test = np_utils.to_categorical(y_test, num_classes=9)
t, v = np.unique(train_data, return_counts=True)
print(t, v)
t, v = np.unique(test_data, return_counts=True)
print(t, v)
return X, X_train, X_test, y_train, y_test
def predict(model, X, X_test, y_test, target_names, classes_authorized, spy_colors, label_dictionary):
classification, confusion, test_loss, test_accuracy = reports(model, X_test, y_test, target_names)
print(classification)
plt.figure(figsize=(13, 10))
plot_confusion_matrix(confusion, classes=target_names,
title='Confusion matrix, without normalization')
X_garbage, train_data, test_data = pp.load_data()
y = np.add(train_data, test_data)
y = pp.delete_useless_classes(y, classes_authorized)
outputs = create_predicted_image(X, y, model, 5, y.shape[0], y.shape[1])
print("PREDICTED IMAGE:")
predict_image = spectral.imshow(classes=outputs.astype(int), figsize=(5, 5))
label_patches = [patches.Patch(color=spy_colors[x] / 255.,
label=label_dictionary[x]) for x in np.unique(y)]
plt.legend(handles=label_patches, ncol=2, fontsize='medium',
loc='upper center', bbox_to_anchor=(0.5, -0.05))
plt.show()
ground_truth = spectral.imshow(classes=y, figsize=(5, 5))
print("IDEAL IMAGE: ")
label_patches = [patches.Patch(color=spy_colors[x] / 255.,
label=label_dictionary[x]) for x in np.unique(y)]
plt.legend(handles=label_patches, ncol=2, fontsize='medium',
loc='upper center', bbox_to_anchor=(0.5, -0.05))
plt.show()
def main():
parser = ArgumentParser(description='Onset Detection Trainer')
parser.add_argument('--network-type',
'-n',
required=True,
choices=['cnn', 'rnn'],
help='network type')
parser.add_argument('--epochs',
type=int,
default=100,
help='number of epochs to train each model for')
req_action = parser.add_mutually_exclusive_group(required=True)
req_action.add_argument('-t',
'--train',
type=int_range,
help='range of models to train')
req_action.add_argument('-e',
'--evaluate',
type=int_range,
help='range of models to evaluate')
args = parser.parse_args()
# Load the eight folds
nn, folds = load_data(args.network_type)
print('* Created folds with sizes %s.' % list(map(len, folds)))
if args.evaluate:
evaluate(nn, folds, args.evaluate)
else:
train(nn, folds, args.train, args.epochs)
def find_xgb_best_parameters(test_size=0.2, n_iter_search=20, X=None, y=None):
if X is None or y is None:
X, y = pr_kaggle.load_data(cat2vectors=True)
Xtrain, Xtest, ytrain, ytest = train_test_split(X,
y,
test_size=test_size,
random_state=36)
param_dist = {
"n_estimators": [50, 100, 250, 500],
"max_depth": [10, 5, 15],
"learning_rate": [0.01, 0.1, 0.0333],
"subsample": [0.5, 1.0, 0.80],
#"gamma": [0,0.01],
#"min_child_weight": [0.5, 1],
"colsample_bytree": [1.0, 0.5, 0.8, 0.9]
}
start = time()
clf = xgb.XGBClassifier()
random_search = RandomizedSearchCV(clf,
param_distributions=param_dist,
n_iter=n_iter_search,
n_jobs=1)
print Xtrain.shape
random_search.fit(Xtrain, ytrain)
print(
"RandomizedSearchCV took %.2f seconds for %d candidates"
" parameter settings." % ((time() - start), n_iter_search))
report(random_search.grid_scores_)
print 'training', random_search.score(Xtrain, ytrain)
print 'testing', random_search.score(Xtest, ytest)
return random_search
def load_all_paths():
train_filenames = ["friday_topknob_bottomknob_switch_slide_0_path",
"friday_microwave_topknob_bottomknob_hinge_0_path",
"friday_microwave_kettle_topknob_switch_0_path",
"friday_microwave_kettle_topknob_hinge_0_path",
"friday_microwave_kettle_switch_slide_0_path",
"friday_microwave_kettle_hinge_slide_0_path",
"friday_microwave_kettle_bottomknob_slide_0_path",
"friday_microwave_kettle_bottomknob_hinge_0_path",
"friday_microwave_bottomknob_switch_slide_0_path",
"friday_microwave_bottomknob_hinge_slide_0_path",
"friday_kettle_topknob_switch_slide_0_path",
"friday_kettle_topknob_bottomknob_slide_1_path",
"friday_kettle_switch_hinge_slide_0_path",
"friday_kettle_bottomknob_switch_slide_0_path",
"friday_kettle_bottomknob_hinge_slide_0_path"
]
#join validation and training data
data_filenames = ["./data/training/%s.pkl"%data_file for data_file in train_filenames]
data_filenames.append("./data/validation/friday_microwave_topknob_bottomknob_slide_0_path.pkl")
#Load data and transform into sequences
all_paths = []
for data_file in data_filenames:
paths = load_data([data_file])[0]
load_idx = list( range(0, paths['images'].shape[0], 3) )#skip 3 frames
for key in paths:
paths[key] = paths[key][load_idx]
all_paths.append(paths)
del paths
return all_paths
def test_load_data(self):
"""
Test splitting of ratings into texts and labels lists
"""
(X_train, y_train), (X_test, y_test) = load_data()
self.assertEqual((25000, ), X_train.shape,
"Incorrect shape of training features")
self.assertEqual((25000, ), X_test.shape,
"Incorrect shape of test features")
self.assertEqual((25000, ), y_train.shape,
"Incorrect shape of training labels")
self.assertEqual((25000, ), y_test.shape,
"Incorrect shape of test labels")
self.assertEqual(list, X_train.dtype,
"Incorrect type of values in training features")
self.assertEqual(list, X_test.dtype,
"Incorrect type of values in test features")
self.assertEqual(np.int64, y_train.dtype,
"Incorrect type of values in training labels")
self.assertEqual(np.int64, y_test.dtype,
"Incorrect type of values in test labels")
self.assertEqual(
set([0, 1]), set(y_train),
"Labels list contains other values than 0 or 1 in trainset")
self.assertEqual(
set([0, 1]), set(y_test),
"Labels list contains other values than 0 or 1 in testset")
def run():
print("Loading data ...")
X_train, Y_train, X_test, submission_file_content = preprocessing.load_data()
print("Performing conversion ...")
X_train = preprocess_images(X_train)
X_test = preprocess_images(X_test)
categorical_Y_train, encoder = preprocess_labels(Y_train)
model = init_model(np.unique(Y_train).size)
if not os.path.isfile(OPTIMAL_MODEL_FILE_PATH):
print("Performing the training phase ...")
if not os.path.isdir(MODEL_FOLDER_PATH):
os.makedirs(MODEL_FOLDER_PATH)
earlystopping_callback = EarlyStopping(patience=1)
modelcheckpoint_callback = ModelCheckpoint(OPTIMAL_MODEL_FILE_PATH, save_best_only=True)
model.fit(X_train, categorical_Y_train, batch_size=BATCH_SIZE, nb_epoch=1,
callbacks=[earlystopping_callback, modelcheckpoint_callback],
validation_split=0.2, show_accuracy=True)
print("Loading the optimal model ...")
model.load_weights(OPTIMAL_MODEL_FILE_PATH)
print("Generating prediction ...")
temp_predictions = model.predict(X_test, batch_size=BATCH_SIZE)
prediction = encoder.inverse_transform(temp_predictions)
print("Writing prediction to disk ...")
submission_file_name = "Aurora_{:.4f}_{:d}.csv".format(EarlyStopping.best, int(time.time()))
submission_file_content[preprocessing.LABEL_COLUMN_NAME_IN_SUBMISSION] = prediction
submission_file_content.to_csv(submission_file_name, index=False)
print("All done!")
def main():
# loads and preprocesses data. See `preprocessing.py`
data, labels, vocabs = load_data(data_dir='./data')
# get embedding
# embedding = get_embedding(vocabs)
# trains a classifier on `train` and `dev` set. See `model.py`
for unit1 in units_1st:
for unit2 in units_2nd:
for em_dim in embedding_dims:
print(
"first layer units, \t second layer units \t embedding dimension"
)
print(unit1, unit2, em_dim)
clf = DRSClassifier(train_labels=labels['train'],
dev_labels=labels['dev'],
vocabs=vocabs,
embedding_dim=em_dim,
unit_1st=unit1,
unit_2nd=unit2)
clf.train(train_instances=data['train'],
dev_instances=data['dev'])
# output model predictions on `test` set
preds_file = "./preds.json"
clf.predict(data['test'], export_file=preds_file)
# measure the accuracy of model predictions using `scorer.py`
run_scorer(preds_file)
def test_gcb(Xy=None, n_estimators=100, max_depth=10, test_size=0.1):
if Xy is None:
X, y = pr_kaggle.load_data()
else:
X, y = Xy
Xtrain, Xtest, ytrain, ytest = train_test_split(X,
y,
test_size=test_size,
random_state=36)
dc = lambda: GBC(
learning_rate=0.02,
n_estimators=n_estimators,
max_depth=max_depth,
# min_samples_split = 4,
# subsample = 0.8, max_features = 0.66
)
clf = dc()
check_classifier(Xtrain, ytrain, Xtest, ytest, clf)
clf = dc()
clfbag = BaggingClassifier(clf, n_estimators=5)
check_classifier(Xtrain, ytrain, Xtest, ytest, clfbag)
clf = dc()
clf_isotonic = CalibratedClassifierCV(clf, cv=5, method='isotonic')
check_classifier(Xtrain, ytrain, Xtest, ytest, clf_isotonic)
def main(unused_argv):
# Load training and eval data
#mnist = tf.contrib.learn.datasets.load_dataset("mnist")
#train_data = mnist.train.images # Returns np.array
#train_labels = np.asarray(mnist.train.labels, dtype=np.int32)
#eval_data = mnist.test.images # Returns np.array
#eval_labels = np.asarray(mnist.test.labels, dtype=np.int32)
train_data, train_labels = load_data(TRAIN_DATA_DIRECTORY)
eval_data, eval_labels = load_data(TEST_DATA_DIRECTORY)
# Create the Estimator
mnist_classifier = tf.estimator.Estimator(
model_fn=cnn_model_fn, model_dir="./tmp/mnist_convnet_model")
# Set up logging for predictions
# Log the values in the "Softmax" tensor with label "probabilities"
tensors_to_log = {"probabilities": "softmax_tensor"}
logging_hook = tf.train.LoggingTensorHook(tensors=tensors_to_log,
every_n_iter=50)
# Train the model
train_input_fn = tf.estimator.inputs.numpy_input_fn(x={"x": train_data},
y=train_labels,
batch_size=100,
num_epochs=None,
shuffle=True)
mnist_classifier.train(input_fn=train_input_fn,
steps=1000,
hooks=[logging_hook])
# Evaluate the model and print results
eval_input_fn = tf.estimator.inputs.numpy_input_fn(x={"x": eval_data},
y=eval_labels,
num_epochs=1,
shuffle=False)
eval_results = mnist_classifier.evaluate(input_fn=eval_input_fn)
print('eval_results: {}'.format(eval_results))
predict_input_fn = tf.estimator.inputs.numpy_input_fn(
x={"x": train_data[2]}, shuffle=False)
prediction_results = mnist_classifier.predict(predict_input_fn)
for i in prediction_results:
print("i: {}".format(i))
print("i['classes']: {}".format(i['classes']))
def main():
args = parse_args()
train_datapath = args.train_dir
model_save_path = args.model_save_dir
label = load_obj(os.path.join("preprocess_file", "label"))
[train_file, train_labels] = load_data(train_datapath, label, False)
k_fold_num = 15
k_fold = cut_CV_data(train_labels, k=k_fold_num) # cross validation K-fold
train_file = np.array(train_file)
train_labels = np.array(train_labels)
if not os.path.exists(model_save_path):
os.makedirs(model_save_path)
for i in range(k_fold_num - 1):
train_f, train_l = np.array([]), np.array([])
for k in range(k_fold_num):
if k != i:
train_f = np.concatenate((train_f, train_file[k_fold[k]]), axis=0)
train_l = np.concatenate((train_l, train_labels[k_fold[k]]), axis=0)
train_dataset = Car196Dataset(
[train_f, train_l], input_transform=my_transform, is_train=True
)
valid_dataset = Car196Dataset(
[train_file[k_fold[i]], train_labels[k_fold[i]]], is_train=False
)
train_loader = DataLoader(
train_dataset, num_workers=4, batch_size=16, shuffle=True
)
valid_loader = DataLoader(
valid_dataset, num_workers=4, batch_size=16, shuffle=False
)
net = torch.hub.load(
"pytorch/vision:v0.6.0", "wide_resnet50_2", pretrained=True
)
net.fc = nn.Linear(2048, 196)
net = net.to(device)
optimizer = optim.SGD(
net.parameters(), lr=0.001, momentum=0.9, weight_decay=0.0045
)
stepLR = optim.lr_scheduler.StepLR(optimizer, 1000, gamma=0.8)
train_early_stop(
net,
train_loader,
valid_loader,
stepLR,
n_steps=1000,
p=6,
savefile=os.path.join(model_save_path, "best_model{}.pt".format(i)),
show_acc=True,
return_log=True,
device=device,
)
def load_paths():
valid_filenames = [
"friday_microwave_topknob_bottomknob_slide_%d_path" % i
for i in [3, 6, 8, 9, 11]
]
data_filenames = [
"./data/validation/%s.pkl" % data_file for data_file in valid_filenames
]
paths = load_data(data_filenames)
return paths
def show_answer():
data = load_data(file_path.get())
Ans = model.predict(data)
for val in Ans:
if val[0] > 0.5:
blank.insert(0, round(val[0] - 0.5 * random.random(), 2))
blank.insert(0, ", ")
else:
blank.insert(0, round(0.5 * random.random(), 2))
blank.insert(0, ", ")
def main():
data = load_data(data_dir='./data')
### Edit any hyperparameters here, including model type. ###
clf = DRSClassifier(model_type='FFN', batches=64, epochs=5)
clf.train(train_instances=[pair[0] for pair in data['train']],
dev_instances=[pair[0] for pair in data['dev']])
preds_file = "./preds.json"
clf.predict(data['test'], export_file=preds_file)
run_scorer(preds_file)
def main():
# X, y, stock_data = load_data()
print("1. Loading data")
X, y, sd = load_data()
print("Finished loading data\n")
print("2. Preprocessing tweets")
processed_X = process_tweets(X)
print("Finished preprocessing tweets\n")
#
print("3. Extracting bag-of-words features")
X_bow_features = bow_build_X(processed_X)
print("Finished extracting bag-of-word features\n")
X_features_stocks = add_stock_feature(X_bow_features, sd)
# #
print("4. Train and test BoW NB")
train_and_test_NB(X_bow_features, y)
print("Finished training and testing NB\n")
# #
print("5. Train and test BoW NB with stocks")
train_and_test_NB(X_features_stocks, y)
print("Finished training and testing NB\n")
# #
print("6. Training BoW models")
train_models(X_bow_features, y, True)
print("Finished training models\n")
print("7. Training BoW models with stocks")
train_models(X_features_stocks, y, True)
print("Finished training models\n")
#
print("8. Extracting word2vec features")
X_tweet2vec_features = tweet2vec_build_X(X)
print("Finished extracting word2vec features\n")
#
X_features_stocks = add_stock_feature(X_tweet2vec_features, sd)
#
print("9. Train and test w2v NB")
train_and_test_NB(X_tweet2vec_features, y)
print("Finished training and testing NB\n")
#
print("10. Train and test w2v NB with stocks")
train_and_test_NB(X_features_stocks, y)
print("Finished training and testing NB\n")
#
print("11. Training w2v models")
train_models(X_tweet2vec_features, y)
print("Finished training models\n")
#
print("12. Training w2v models with stocks")
train_models(X_features_stocks, y)
print("Finished training models\n")
def app():
st.title('Finder')
step1 = st.button('Did you get the Coordinate?', key=1)
step2 = st.button('Checking the Coordinate', key=2)
step3 = st.button('Processing', key=3)
step4 = st.button('Finding Place of ..', key=4)
if step1:
st.markdown(
""
)
st.markdown(
"<h3 style='text-align: center; color: red;'>Coordinate : 9.16085726217318 <------> 8.807258629151487</h3>",
unsafe_allow_html=True,
)
if step2:
st.markdown(
""
)
data = load_data("planet")
st.subheader('Look at the Data')
data_load_state = st.text('Loading data...')
st.write(data.head(10))
# Notify the reader that the data was successfully loaded.
data_load_state.text('Loading data...done!')
if step3:
st.header('Procedures')
st.markdown('* Find the point form locations list')
st.markdown(
'* Creating model, fit it, finding mean values of coordinates')
st.markdown('* Show the location in map ')
st.text(' ')
st.subheader("Locations in the map")
image = Image.open('./graphics/map.png')
st.image(image, caption="Yoda is at one of them ")
st.text(' ')
st.markdown(
""
)
if step4:
st.header('Finally')
st.markdown('* Where is the little Baby ')
st.text(' ')
image = Image.open('./graphics/pla.png')
st.image(image, caption="Do you see ?")
st.text(' ')
st.markdown(
""
)
st.balloons()
def load_data(limit=0, split=0.8):
"""Load data from the IMDB dataset."""
# Partition off part of the train data for evaluation
#train_data, _ = thinc.extra.datasets.imdb()
train_data = preprocessing.load_data(include_body=False)
random.shuffle(train_data)
train_data = train_data[-limit:]
texts, labels = zip(*train_data) #a tuple(iterable) full of text and a tuple full of labels
cats = [{'POSITIVE': bool(y)} for y in labels]
split = int(len(train_data) * split)
return (texts[:split], cats[:split]), (texts[split:], cats[split:])
def _pipeline_bow_and_multinomialNB(self, dir_dataset):
categories = os.listdir(dir_dataset)
# X_train
X_train, y_train = [], []
for category in tqdm(categories):
path = os.path.join(dir_dataset, category, 'evc.train.en')
X_tmp, y_tmp = load_data(path)
for i in range(len(X_tmp)):
X_train.append(X_tmp[i])
y_train.append(y_tmp[i])
# X_test
X_test, y_test = [], []
for category in tqdm(categories):
path = os.path.join(dir_dataset, category, 'evc.test.en')
X_tmp, y_tmp = load_data(path)
for i in range(len(X_tmp)):
X_test.append(X_tmp[i])
y_test.append(y_tmp[i])
# transform text to vector by word of bag
BoW = WordEmbedding().CountVectorizer()
X_train = BoW.fit_transform(X_train)
X_train = X_train.toarray()
print(X_train.shape, len(y_train))
X_test = BoW.transform(X_test).toarray()
model = MultinomialNB_custom()
model.fit(X_train, y_train)
y_hat = model.predict(X_test)
print(classification_report(y_test, y_hat))
return BoW, model
def main():
# Load dataset by uncommenting any of the required dataset.
# Run the inputDataCollector first if you want to uncomment any of the below dataset
# data = pp.load_data('Dataset/stocknews/RedditNews.csv')
# data = pp.load_data('Dataset/tweets.csv')
# data = pp.load_data('Dataset/news.csv')
data = pp.load_data('Dataset/news.csv')
data = pd.DataFrame(data)
# This line is to concatenate the title & summary in a news articles for the export dataset
text = list(data["TITLE"] + data["SUMMARY"])
# Uncomment the below line for all the other datasets
# text = list(data["Text"])
# calculate the labels for the task
labels = np.array(list(data["Label"]))
# preprocess the dataset
text = pp.preprocess(text)
#tokenize the text
tokenizer = Tokenizer(num_words=10000)
tokenizer.fit_on_texts(text)
sequences = tokenizer.texts_to_sequences(text)
word_index = tokenizer.word_index
# padding sequences to ensure all rows are of equal length
max_review_length = 1000
text = pad_sequences(sequences,
maxlen=max_review_length,
padding='pre',
truncating='pre')
#split the data into 80-20 train-test
X_train = text[:int(0.8 * len(text))]
X_val = text[-int(0.2 * len(text)):]
y_train = labels[:int(0.8 * len(text))]
y_val = labels[-int(0.2 * len(text)):]
y_train = [int(i) for i in y_train]
y_val = [int(i) for i in y_val]
y_train = np.array(y_train)
y_val = np.array(y_val)
# build the model & validate the results
nn.build(word_index, X_train, y_train, X_val, y_val)
def main(create_sub=False):
## Load the data
train_df, test_df = pp.load_data()
## Lowercase all the text in the tweets
train_df = pp.lowercase_df(train_df)
test_df = pp.lowercase_df(test_df)
nlp = spacy.load('en_core_web_lg')
# vecs = create_word_vectors(nlp, 'linear_vectorizer_lower_case_hashtag', output=True, disable_pipes=False)
vecs = load_word_vectors('linear_vectorizer_lower_case',
'training_word_vectors.npy')
X_train, X_val, y_train, y_val = train_test_split(vecs,
train_df['target'],
test_size=0.1,
random_state=1)
## Hyper-parameter optimization
# Grid_search_CV(X_train, y_train)
# best_params = Radial_SVM_Random_search_CV(X_train, y_train, 40)
## Fit and run models
model_name = "RBF_SVM_lowercase"
# The best parameters are {'gamma': 0.001, 'C': 10000.0} with a score of 0.77
# rbf_svm, valid_preds = radial_SVM(X_train, X_val, y_train, y_val, {'gamma': 0.1, 'C': 10.0})
# save_model(rbf_svm, "RBF_SVM_lowercase")
rbf_svm = load_model(model_name)
predictions = rbf_svm.predict(X_val)
# polynomial_SVM(X_train, X_val, y_train, y_val)
## Save the model and generate the performance report.
model_dir = os.path.join(MODEL_DIR, model_name.title())
model_eval.generate_model_report(model_name, model_dir, y_val, predictions)
if create_sub == True:
# Apply model on Test data set for submission
with nlp.disable_pipes():
test_vectors = np.array([
nlp(tweet['text']).vector for idx, tweet in test_df.iterrows()
])
# nlp.add_pipe(hashtag_pipe)
# test_vectors = np.array([nlp(tweet['text']).vector for idx, tweet in train_df.iterrows()])
preds_test = rbf_svm.predict(test_vectors)
print(preds_test)
create_submission("rbf_lowercase_submission.csv", preds_test, test_df)
def load_data():
x, y, vocabulary, vocabulary_inv_list = preprocessing.load_data()
vocabulary_inv = {key: value for key, value in enumerate(vocabulary_inv_list)}
# Shuffle data
shuffle_indices = np.random.permutation(np.arange(len(y)))
x = x[shuffle_indices]
y = y[shuffle_indices]
train_len = int(len(x) * 0.9)
x_train = x[:train_len]
y_train = y[:train_len]
x_test = x[train_len:]
y_test = y[train_len:]
return x_train, y_train, x_test, y_test, vocabulary_inv
def main():
# loads and preprocesses data. See `preprocessing.py`
data = load_data(data_dir='./data')
# trains a classifier on `train` and `dev` set. See `model.py`
clf = DRSClassifier()
clf.train(train_instances=data['train'], dev_instances=data['dev'])
# output model predictions on `test` set
preds_file = "./preds.json"
clf.predict(data['test'], export_file=preds_file)
#print(preds)
# measure the accuracy of model predictions using `scorer.py`
run_scorer(preds_file)
def sentiment_analysis(dataset):
if dataset == train_set and os.path.isfile(sentiment_train_pkl):
return load_sentiment(sentiment_train_pkl)
if dataset == val_set and os.path.isfile(sentiment_val_pkl):
return load_sentiment(sentiment_val_pkl)
nltk.download('vader_lexicon', quiet=True)
# load data from csv
data_original = load_data(dataset)
# print(data_original)
#Only go through the first 10 entries of dataset - Remove for entire dataset
# data_original = data_original.head(20)
sid = SentimentIntensityAnalyzer()
sentiment_score = pd.DataFrame(columns=['compound', 'neg', 'neu', 'pos'])
story_idx = 0
#iterate through dataframe for sentiment analysis
for index, row in data_original.iterrows():
#print(row)
story_to_complete = " ".join(
[row['sen1'], row['sen2'], row['sen3'], row['sen4']])
#story_to_complete = "'''{0}'''".format(story_to_complete)
# print(story_to_complete)
scores = sid.polarity_scores(story_to_complete)
story_idx = story_idx + 1
if (story_idx % 10000 == 0):
print(story_idx, "/", data_original.shape[0])
for key in sorted(scores):
# print('{0}:{1}, '.format(key, scores[key]), end='')
#print(scores[key])
sentiment_score.loc[index] = scores
if dataset == train_set:
with open(sentiment_train_pkl, 'wb') as output:
pickle.dump(sentiment_score, output, pickle.HIGHEST_PROTOCOL)
elif dataset == val_set:
with open(sentiment_val_pkl, 'wb') as output:
pickle.dump(sentiment_score, output, pickle.HIGHEST_PROTOCOL)
return sentiment_score
def run():
print("Resetting the submission folder {:s} ...".format(os.path.basename(submission_folder_path)))
shutil.rmtree(submission_folder_path, ignore_errors=True)
os.makedirs(submission_folder_path)
print("Loading data ...")
X_train, Y_train, X_test, submission_file_content = preprocessing.load_data()
print("Tuning parameters ...")
optimal_max_depth, optimal_min_child_weight, optimal_subsample, optimal_colsample_bytree = perform_tuning(X_train, Y_train)
print("Training ...")
optimal_learning_rate = 0.05
estimator = XGBClassifier(max_depth=optimal_max_depth, learning_rate=optimal_learning_rate, n_estimators=1000000,
min_child_weight=optimal_min_child_weight, subsample=optimal_subsample,
colsample_bytree=optimal_colsample_bytree, objective=OBJECTIVE)
generate_prediction(estimator, X_train, Y_train, X_test, submission_file_content, early_stopping_rounds=200, cv_num=20)
print("All done!")
def create_model(file_path=FINAL_MLKNN_MODEL_FILE_PATH):
"""
Creates and trains a MLkNN classifier using the optimized parameters found
Saves this trained model to disk
:param string file_path: specifies where the model should be saved
:return: a trained sklearn MLkNN classifier
"""
with open(OPTIMIZED_MODEL_PARAMETERS_FILE_PATH) as file:
hyperparameters = json.load(file)['hyperparameters']
question_data, music_data = preprocessing.load_data()
question_data, music_data = preprocessing.preprocess_data(
question_data, music_data)
clf = MLkNN(k=hyperparameters['k'], s=hyperparameters['s'])
clf.fit(question_data.values, music_data.values)
pickle.dump(clf, open(file_path, 'wb'))
return clf
def create_model(file_path=FINAL_XGBOOST_MODEL_FILE_PATH):
"""
Creates and trains a OneVsRestClassifier(XGBClassifier()) using the optimized parameters found
Saves this trained model to disk
:param string file_path: specifies where the model should be saved
:return: a trained OneVsRestClassifier
"""
with open(OPTIMIZED_MODEL_PARAMETERS_FILE_PATH) as file:
hyperparameters = json.load(file)['hyperparameters']
question_data, music_data = preprocessing.load_data()
question_data, music_data = preprocessing.preprocess_data(
question_data, music_data)
xgb_model = XGBClassifier(**hyperparameters)
xgb_clf = OneVsRestClassifier(xgb_model, n_jobs=-1)
xgb_clf.fit(question_data, music_data)
pickle.dump(xgb_clf, open(file_path, 'wb'))
return xgb_clf
def main():
args = parse_args()
test_datapath = args.test_dir
model_dir = args.model_dir
save_name = args.save_name
num2name = load_obj(os.path.join("preprocess_file", "num_to_name"))
test_dataset = Car196Dataset(load_data(test_datapath, clean=False), is_train=False)
test_loader = DataLoader(test_dataset, num_workers=4, batch_size=16, shuffle=False)
net = torch.hub.load("pytorch/vision:v0.6.0", "wide_resnet50_2", pretrained=False)
net.fc = nn.Linear(2048, 196)
net = net.to(device)
path_pattern = model_dir + "/**/*.*"
files_list = glob.glob(path_pattern, recursive=True)
csv_list = [["id", "label"]]
tmp = 0
for inputs, labels in test_loader:
inputs = inputs.to(device)
outputs = torch.zeros(
inputs.shape[0], 196, dtype=torch.float64, device=device
).data
for file_name in files_list:
net.load_state_dict(torch.load(file_name))
net.eval()
outputs += nn.Softmax(dim=1)(net(inputs)).data
_, preds = outputs.max(1)
for index, pred in enumerate(preds):
input_file = test_dataset.image_filenames[tmp + index]
id_ = int(basename(input_file).split(".")[0])
csv_list.append([id_, num2name[pred.item()]])
tmp += index + 1
if not os.path.exists("result"):
os.makedirs("result")
with open(os.path.join("result", save_name), "w", newline="") as f:
writer = csv.writer(f)
writer.writerows(csv_list)
def resume_model():
x, y, vocabulary, vocabulary_inv = pre.load_data()
# Randomly shuffle data
shuffle_indices = np.random.permutation(np.arange(len(y)))
x_shuffled = x[shuffle_indices]
y_shuffled = y[shuffle_indices]
x_train, x_val = x_shuffled[:-1000], x_shuffled[-1000:]
sess = tf.Session()
cnn = SentenceCNN(
sequence_length=x_train.shape[1],
num_classes=2,
vocab_size=len(vocabulary),
sess=sess
)
cnn.inference()
cnn.train()
# Create a saver.
saver = tf.train.Saver()
checkpoint_dir = os.path.abspath(os.path.join(config.out_dir, "checkpoints"))
checkpoint_prefix = os.path.join(checkpoint_dir, "model")
ckpt = tf.train.get_checkpoint_state(checkpoint_dir)
if ckpt and ckpt.model_checkpoint_path:
# Restores from checkpoint
saver.restore(sess, ckpt.model_checkpoint_path)
cnn.sess = sess
# Assuming model_checkpoint_path looks something like:
# /my-favorite-path/cifar10_train/model.ckpt-0,
# extract global_step from it.
global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
return cnn
else:
print('No checkpoint file found. Cannot resume.')
return None
def run_sparse_autoencoder(N, image_size, patch_size, prepare_data=True):
# images_all, y, images_repr = prepare_data()
# open training data
print "Trainig data!"
train_path = 'data/train_32x32.mat'
test_path = 'data/test_32x32.mat'
file_train = "data/pickles/train.pickle"
file_val = "data/pickles/val.pickle"
if prepare_data:
X, y = prepr.load_data(train_path)
X_test, y_test = prepr.load_data(test_path)
prepr.normalize_and_pickle(X, y, X_test, y_test)
print "Training data were loaded and normalized!"
images_train = helper.unpickle_data(file_train)[:, :, :N]
images_val = helper.unpickle_data(file_val)
images_repr = images_val[:, :, :36]
# theta = init.initialize_k_deep_sparse_autoencoder(patch_size, image_size)
theta = init.init_original_model(image_size)
# max_iter = 5
# batch_size = 1000
# n_batches = N // batch_size
# print "n_batches: ", n_batches
# learning_rate = 1e-3
# learning_rate_decay = 0.95
# mu = 0.9
lambda_ = 0.001
# iter = 0
# v = {}
# whole_loss_history = []
# train_loss_history = []
# val_loss_history = []
# while iter < max_iter:
# iter += 1
# s = 0
# for b in range(n_batches):
# batch_begin = b * batch_size
# N_average = (b + 1) * batch_size
# batch_end = batch_begin + batch_size
# X_batch = images_train[:, :, batch_begin:batch_end]
# cost, grad = model.k_sparse_deep_autoencoder_cost_without_patches(theta, lambda_, X_batch, patch_size, image_size, batch_size, patch_size)
# whole_loss_history.append(cost)
# # momentum update
# for item in grad:
# if item not in v:
# v[item] = np.zeros(grad[item].shape)
# v[item] = mu * v[item] - learning_rate * grad[item]
# theta[item] += v[item]
# mask = np.random.choice(N, 1000)
# train_subset = images_train[:, :, mask]
# cost_train = model.k_sparse_deep_autoencoder_cost_without_patches(theta, lambda_, train_subset, patch_size, image_size, 1000, patch_size)[0]
# train_loss_history.append(cost_train)
# cost_val = model.k_sparse_deep_autoencoder_cost_without_patches(theta, lambda_, images_val, patch_size, image_size, images_val.shape[2], patch_size)[0]
# val_loss_history.append(cost_val)
# print "Cost_train: ", cost_train, ", cost_val: ", cost_val, ", epoch: ", iter, " learning_rate: %d", (learning_rate)
# learning_rate *= learning_rate_decay
# print "Check gradients!"
# lambda_ = 0.1
l_cost, l_grad = original_model.k_sparse_original_model(theta, lambda_, images_train, patch_size, image_size, N,
15)
# helper.check_sparsity_of_gradients(l_grad, 'W3')
# J = lambda x: model.k_sparse_deep_autoencoder_cost_without_patches(x, lambda_, images_train, patch_size, image_size, N, 2)
# gradient_check.compute_grad(J, theta, l_grad)
J = lambda x: original_model.k_sparse_original_model(x, lambda_, images_train, patch_size, image_size, N, 15)
gradient_check.compute_grad(J, theta, l_grad)
# Returns 10 movie recommendations given user's age, gender and occupation
import numpy as np
import preprocessing
import testing
from npkmeans import kmeans
# Run k means
km = kmeans(preprocessing.load_data("u.data", 1), 100)
kmeans_return = km.kCluster()
clusters = kmeans_return[0]
centroids = kmeans_return[1]
testing.calculate(clusters, centroids)
def sim_users_ratings(age, gender, occupation):
"Calculates similar users, finds those users' centroids, averages centroid ratings and returns a dictionary of sorted movie indexes"
# Load users data
users = preprocessing.user_load_data("u.user")
sim_users = []
# Iterate through users, adding indexes of users with same age, gender and occupation
for position, user in enumerate(users):
if (age, gender, occupation) == (user[0], user[1], user[2]):
sim_users.append(position)
# Check whether sim_users is empty, if it is, then....
if len(sim_users) == 0:
for position, user in enumerate(users):
if (age, occupation) == (user[0], user[2]):
sim_users.append(position)
print ("No exact match found, matching for age and occupation...")
tf.flags.DEFINE_string("checkpoint_dir", "", "Checkpoint directory from training run")
# Misc params
tf.flags.DEFINE_boolean("allow_soft_placement", True, "Allow soft device placement")
tf.flags.DEFINE_boolean("log_device_placement", False, "log placement of ops on devices")
FLAGS = tf.flags.FLAGS
FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(FLAGS.__flags.items()):
print("{}={}".format(attr.upper(), value))
print("")
# Load my own test data here
_, _, vocab, vocab_inv = preprocessing.load_data(is_train=True)
x_test, y_test, _, _ = preprocessing.load_data(is_train=False)
y_test = np.argmax(y_test, axis=1)
print("Vocabulary size: {:d}".format(len(vocab)))
print("Test set size: {:d}".format(len(y_test)))
print("\nEvaluating on test data..\n")
checkpoint_file = tf.train.latest_checkpoint(FLAGS.checkpoint_dir)
graph = tf.Graph()
with graph.as_default():
session_conf = tf.ConfigProto(
allow_soft_placement=FLAGS.allow_soft_placement,
log_device_placement=FLAGS.log_device_placement)
sess = tf.Session(config=session_conf)
with sess.as_default():
lambda_l2_param_search = [0.001, 0.01, 0.1, 1, 10]
embedding_dim_param_search = [64, 128, 256]
tf.flags.FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(tf.flags.FLAGS.__flags.items()):
print("{}={}".format(attr.upper(), value))
print("")
###########################################
# Data Preparation
###########################################
print("loading data...")
x, y, vocab, vocab_inv = preprocessing.load_data()
# Shuffle data randomly
np.random.seed(10)
# shuffled_indices = np.random.permutation(np.arange(len(y)))
# x_shuffled = x[shuffled_indices]
# y_shuffled = y[shuffled_indices]
# split train/valid set
# TODO Implement a f*****g correct XVal procedure for this
x_train, x_valid = x[:-500], x[-500:]
y_train, y_valid = y[:-500], y[-500:]
print("Vocabulary Size: {:d}".format(len(vocab)))
print("Train/Validation split: {:d}/{:d}".format(len(y_train), len(y_valid)))
###########################################
