from parse import image_parse, label_parse
from sklearn import datasets, neighbors, linear_model, metrics
from sklearn.neural_network import MLPClassifier
from joblib import dump
import matplotlib.pyplot as plt
import cv2
digits = image_parse("train-images-idx3-ubyte").reshape(60000, 784)
for pic in digits:
for pixel in pic:
if pixel > 0:
pixel = 255
X_train = (255 - digits) / 255.0
y_train = label_parse("train-labels-idx1-ubyte")
X_test = image_parse("t10k-images-idx3-ubyte").reshape(10000, 784)
X_test = (255 - X_test) / 255.0
y_test = label_parse("t10k-labels-idx1-ubyte")
clf = MLPClassifier(solver='sgd',
alpha=1e-5,
hidden_layer_sizes=(1500, ),
random_state=1)
clf.fit(X_train, y_train)
score = clf.score(X_test, y_test)
print(score)
ax = []
fig = plt.figure(figsize=(20, 20))
from parse import image_parse, label_parse
from sklearn import datasets, neighbors, linear_model, metrics
from sklearn.neighbors import KNeighborsClassifier
from joblib import dump, load
import matplotlib.pyplot as plt
X_test = image_parse("t10k-images-idx3-ubyte").reshape(10000, 784)
X_test = (255 - X_test) / 255.0
y_test = label_parse("t10k-labels-idx1-ubyte")
knn = load("knn.joblib")
#score = knn.score(X_test, y_test)
#print(score)
ax = []
fig = plt.figure(figsize=(20, 20))
for num in range(100):
img = X_test[num].reshape(28, 28)
ax.append(fig.add_subplot(10, 10, num + 1))
ax[-1].set_title(
str(knn.predict(X_test[num].reshape(1, -1))) + " " + str(y_test[num]))
ax[-1].axis('off')
plt.imshow(img, cmap=plt.cm.gray_r, interpolation='nearest')
fig.tight_layout()
plt.axis('off')
plt.show()