def save():
data = get_data()
model = KNN()
model.fit(data)
pickle.dump(model, open('model.pkl', 'wb'))
def predict_knn(filepath, cells):
knn = KNN(3, train=False)
knn.load_knn(filepath)
grid = ""
for cell in cells:
cell = cell.reshape(1, -1)
pred = knn.predict(cell)
# period character was used as the wildcard value, but on the backend
# the model is trained to predict a 0, so it is remapped.
if pred == 0:
pred = "."
grid += str(pred)
return grid
def compare_models(hidden_dim, iterations, input_dim = 3, verbose = False):
f1 = Symmetric(input_dim, hidden_dim, hidden_dim, 10)
f2 = KNN(input_dim, hidden_dim, hidden_dim, 10)
f3 = KK(input_dim, hidden_dim, hidden_dim, 10)
f1.__name__ = "S1"
f2.__name__ = "S2"
f3.__name__ = "S3"
models = [f1, f2, f3]
lambs = [0., 1e-6, 1e-4, 1e-2]
for model in models:
print("model", model.__name__)
cv_models = cross_validate(model, train_loader, iterations, lambs, verbose)
validation_errors = np.zeros_like(lambs)
for i, cv_model in enumerate(cv_models):
validation_errors[i] = test(cv_model, train_loader)
i = np.argmin(validation_errors)
lamb = lambs[i]
runs = 3
run_errors = np.zeros(runs)
for i in range(runs):
print("run", i)
model_copy = copy.deepcopy(model)
model_copy.reinit()
train(model_copy, train_loader, iterations, lamb)
run_errors[i] = test(model_copy, test_loader)
mean_error = np.mean(run_errors)
std_error = np.std(run_errors)
print("mean: {}, std: {}".format(mean_error, std_error))
def main():
k = 3
split = 0.8
header, x_train, y_train, x_test, y_test = KNN.load('IRIS.csv', split)
knn = KNN(x_train, y_train, k)
y_pred = knn.test(x_test)
pprint(y_pred)
accuracy = KNN.accuracy(y_pred, y_test)
print(f"Acuracy est de {accuracy}")
flower = [6.1, 2.9, 4.7, 1.4]
prediction = knn.test([flower])
print(f"fleur {flower} est un {prediction}")
def main():
k = 3
split = 0.8
header, x_train, y_train, x_test, y_test = KNN.load('iris.csv', split)
knn = KNN(x_train, y_train, k)
y_pred = knn.test(x_test)
pprint(y_pred)
accuracy = KNN.accuracy(y_pred, y_test)
print(f"Accuracy est de: {accuracy}")
flower = [0.8, 0.0, 5, 0]
prediction = knn.test([flower])
print(f"La fleur {flower} est un {prediction[0]}")
train_sampler = SubsetRandomSampler(train_idx)
val_sampler = SubsetRandomSampler(valid_idx)
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=args.batch_size, sampler = train_sampler)
val_loader = torch.utils.data.DataLoader(train_dataset, batch_size=args.batch_size, sampler = val_sampler)
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=args.batch_size)
input_dim = 3
if args.model == 'overkill':
model = Overkill(input_dim, args.h1, args.h2, args.h3, 10).to(device)
elif args.model == 's1':
model = Symmetric(input_dim, args.h1, args.h2, 10).to(device)
elif args.model == 's2':
model = KNN(input_dim, args.h1, args.h2, 10).to(device)
elif args.model == 's3':
model = KK(input_dim, args.h1, args.h2, 10).to(device)
train(model, train_loader, args.iterations, lamb = args.lamb, lr = args.lr)
print(args)
error = test(model, train_loader)
print("train error: ", error)
error = test(model, val_loader)
print("val error: ", error)
error = test(model, test_loader)
print("test error: ", error)
@author: Avinash
"""
from model import KNN
from utils import get_mnist
mnist_data=get_mnist()
X=mnist_data[:,1:]
Y=mnist_data[:,0]
X_train=X[0:int(X.shape[0]*0.9)]
Y_train=Y[0:int(X.shape[0]*0.9)]
X_test=X[int(X.shape[0]*0.9):]
Y_test=Y[int(X.shape[0]*0.9):]
#X_train/=255.0
model=KNN()
model.fit(X_train,Y_train,n=5)
#X_test/=255.0
test_acc=model.evaluate(X_test,Y_test)
print("Test Accuracy : ",test_acc)
'''
Test Accuracy : 0.9671428571428572
'''
prompt += f" {next_}"
print()
def list_predictions(model: Model,
prompt: str,
num_predictions: int = 5) -> None:
token_odds = model.predict(prompt)
for i, item in enumerate(list(token_odds.items())):
print(f"\t{i + 1}. '{item[0]}' ({item[1]})")
if i + 1 == num_predictions:
return
prompt = "The hitch hiker's guide to the"
if __name__ == "__main__":
print(f"Prompt: {prompt}")
print("NAIVE")
m: Model = Naive(2, history=5)
m.fit(text)
list_predictions(m, prompt)
free_write(m, prompt)
print("KNN")
m = KNN(2, {Distance.NAIVE, Distance.WU_PALMER})
m.fit(text)
list_predictions(m, prompt)
free_write(m, prompt)
# -*- encoding: utf-8 -*-
"""
@Description: 实现了原始的KNN，损失函数默认使用欧氏距离。
@Time : 2021-3-13 21:03
@File : main.py
@Software: PyCharm
"""
from model import KNN
from utils import setup_seed, dataloader
import numpy as np
from sklearn.metrics import accuracy_score
seed = 1234
setup_seed(seed)
x_train, x_test, y_train, y_test = dataloader(ratio=0.9)
knn = KNN(k=5)
test_num = x_test.shape[0]
# train
knn.train(instences=x_train, labels=y_train)
# test
y_pred_list = list()
for instance_id in range(test_num):
x = x_test[instance_id]
y_pred = knn.eval(x)
y_pred_list.append(y_pred)
y_pred_array = np.array(y_pred_list)
print("Final Accuracy: {}%.".format(
accuracy_score(y_pred_array, y_test) * 100))