def main():
print("Reading dataset...")
X_train, Y_train, Y_train_raw = read_train()
print("Dataset ready.")
print("Start training...")
learning_rate = 0.045
num_iterations = 2500
parameters, costs = nn_model(X_train,
Y_train,
num_iterations=num_iterations,
learning_rate=learning_rate,
print_cost=True)
print("Traning finished.")
print("Predicting train set...")
Y_prediction_train = predict(X_train, parameters)
precision, recall, fscore, _ = precision_recall_fscore_support(
Y_train_raw.reshape((X_train.shape[1], 1)),
Y_prediction_train,
labels=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
print("Precistion:", precision)
print("Recall:", recall)
print("Fscore:", fscore)
print("Reading test set...")
X_test = read_test()
print("Predicting test set...")
Y_prediction_test = predict(X_test, parameters)
print("Predicted.")
costs = np.squeeze(costs)
plt.plot(costs)
plt.ylabel("cost")
plt.xlabel("iterations")
plt.title("Learning rate = " + str(learning_rate))
plt.show()
d = {
"costs": costs.tolist(),
"Y_prediction_test": Y_prediction_test.tolist(),
"Y_prediction_train": Y_prediction_train.tolist(),
"W1": parameters["W1"].tolist(),
"W2": parameters["W2"].tolist(),
"b1": parameters["b1"].tolist(),
"b2": parameters["b2"].tolist(),
"learning_rate": learning_rate,
"num_iterations": num_iterations
}
print("Saving model...")
jd = json.dumps(d)
with open("model.json", "w") as f:
f.write(jd)
print("Model saved.")
def classifyRF(train_file="train.csv", test_file ="test.csv", trees=70):
#So, let's classifiy this thing. Reading the Features and then the test.
print("Reading train data")
X,y = rd.read_train(file_name=train_file)
print("Augmenting dataset")
X,y = rd.nudge_dataset(X,y)
print("Reading test data")
test_data = rd.read_test(file_name=test_file)
#Creating the classifier. It has a ton of parameters, but since this a hard and fast one, here you go.
rfc = RandomForestClassifier(trees)
#Train with the data we have. Cry a little inside.
print("Training classifier")
rfc.fit(X, y)
predictions = rfc.predict(test_data)
#Most submitions are cute with a CSV. Might as well learn how to do it.
pd.DataFrame({"ImageId": range(1,len(predictions)+1), "Label": predictions}).to_csv('submit.csv', index=False, header=True)
def classifyRF(train_file="train.csv", test_file="test.csv", trees=70):
#So, let's classifiy this thing. Reading the Features and then the test.
print("Reading train data")
X, y = rd.read_train(file_name=train_file)
print("Augmenting dataset")
X, y = rd.nudge_dataset(X, y)
print("Reading test data")
test_data = rd.read_test(file_name=test_file)
#Creating the classifier. It has a ton of parameters, but since this a hard and fast one, here you go.
rfc = RandomForestClassifier(trees)
#Train with the data we have. Cry a little inside.
print("Training classifier")
rfc.fit(X, y)
predictions = rfc.predict(test_data)
#Most submitions are cute with a CSV. Might as well learn how to do it.
pd.DataFrame({
"ImageId": range(1,
len(predictions) + 1),
"Label": predictions
}).to_csv('submit.csv', index=False, header=True)
def get_train_and_test_data(train_limit=-1,test_limit=-1):
X,y = get_train_data(train_limit)
print('Loading test data')
test_X = rd.read_test(limit=test_limit)
test_X = scale(test_X)
return X,y,test_X
def get_train_and_test_data(train_limit=-1, test_limit=-1):
X, y = get_train_data(train_limit)
print('Loading test data')
test_X = rd.read_test(limit=test_limit)
test_X = scale(test_X)
return X, y, test_X