To answer this SO question: https://stackoverflow.com/questions/56445760/do-i-need-to-extract-feature-vectors-from-mnist-before-using-kmeans/56445967?noredirect=1#comment99505544_56445967

KNN over MNIST

from sklearn.datasets import load_digits
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
import numpy as np

digits = load_digits()
x = list(map(lambda row: row.flatten() / 255, digits.data))
y = digits.target

x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.33, random_state=42)

model = KNeighborsClassifier(n_neighbors=3)
model.fit(x_train, y_train)

y_train_prediction = model.predict(x_train)
y_test_prediction = model.predict(x_test)

train_accuracy = accuracy_score(y_train, y_train_prediction)
test_accuracy = accuracy_score(y_test, y_test_prediction)
train_confusion = confusion_matrix(y_train, y_train_prediction, labels=np.arange(0, 10))
test_confusion = confusion_matrix(y_test, y_test_prediction, labels=np.arange(0, 10))

print('train accuracy: {}% | test accuracy: {}%'.format(train_accuracy * 100, test_accuracy * 100))


def print_matrix(mat):
    print(' ', end=' ')
    for i in range(10):
        print(str(i).ljust(3), end=' ')
    print()
    for i in range(10):
        print(i, end=' ')
        for j in range(10):
            print(str(mat[i, j]).ljust(3), end=' ')
        print()


print('===')
print('train:')
print_matrix(train_confusion)
print('test:')
print_matrix(test_confusion)