Day 05_Multivariate Gaussian_Winery_Classifier_MNIST

# -*- coding: utf-8 -*-
"""Day 09_Multivariable Gaussian Classifier_MNIST.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1yQhgjwD07GjN_vtmmM7B4sF-I_92ChOU
"""
 
from google.colab import drive
drive.mount('/gdrive')
 
PATH = "/gdrive/My Drive/Colab Notebooks/resources/"
 
# %matplotlib inline
import matplotlib.pyplot as plt 
import gzip, os
import numpy as np
from scipy.stats import multivariate_normal
import sys
 
if sys.version_info[0] == 2:
    from urllib import urlretrieve
else:
    from urllib.request import urlretrieve
 
# Function that downloads a specified MNIST data file from Yann Le Cun's website
def download(filename, source='http://yann.lecun.com/exdb/mnist/'):
    print("Downloading %s" % filename)
    urlretrieve(source + filename, filename)
 
# Invokes download() if necessary, then reads in images
def load_mnist_images(filename):
    if not os.path.exists(filename):
        download(filename)
    with gzip.open(filename, 'rb') as f:
        data = np.frombuffer(f.read(), np.uint8, offset=16)
    data = data.reshape(-1,784)
    return data
 
def load_mnist_labels(filename):
    if not os.path.exists(filename):
        download(filename)
    with gzip.open(filename, 'rb') as f:
        data = np.frombuffer(f.read(), np.uint8, offset=8)
    return data
 
def displaychar(image):
    plt.imshow(np.reshape(image, (28,28)), cmap=plt.cm.gray)
    plt.axis('off')
    plt.show()
 
## Load the training set
train_data = load_mnist_images('train-images-idx3-ubyte.gz')
train_labels = load_mnist_labels('train-labels-idx1-ubyte.gz')
 
## Load the testing set
test_data = load_mnist_images('t10k-images-idx3-ubyte.gz')
test_labels = load_mnist_labels('t10k-labels-idx1-ubyte.gz')
 
train_data.shape , train_labels.shape
 
displaychar(train_data[0])
 
def fit_generative_model(x,y):
    
    #사진 타겟 클래스의 갯수 
    k = 10
    
    #784 = 셀의 갯수
    d = (x.shape)[1]
    
    #피처 각각의 평균 0번은 쓰지 않고 1, 2, 3번 인덱스만 사용.
    mu = np.zeros((k,d))
    
    #피처 각각의 표준편차
    cov_mat = np.zeros((k,d,d))
    
    #클래스 분류시 가중치로 쓰일 pi변수 1, 2, 3번 인덱스만 사용.
    pi = np.zeros(k)
    
    
    #타겟 클래스의 수만큼 도는 포문
    for label in range(0,k):
        
        #타겟 클래스가 label과 일치하는 인덱스 추출. ( 1~3 )
        indices = (y == label)
        
        #해당 클래스의 평균을 mu의 클래스 번호 인덱스에 할당.
        mu[label] = np.mean(x[indices,:], axis=0)
        
        #해당 클래스의 공분산 매트릭스를 mu의 클래스 번호 인덱스에 할당.
        cov_mat[label] = np.cov(x[indices,:], rowvar=0, bias=1)
        
        #전체 y의 수에서 해당 타겟 클래스의 갯수를 pi의 타겟클래스 인덱스에 저장. 나중에 가중치로 쓰일 예정
        pi[label] = float(sum(indices))/float(len(y))
    return mu, cov_mat, pi
 
mu, cov_mat, pi = fit_generative_model( train_data, train_labels )
 
## regularize
 
## c is a magic number
c = 4
 
for i in range(10):
    eyemat = np.eye(cov_mat[i]. shape[0], cov_mat[i]. shape[1]) * c
    cov_mat[i] += eyemat
 
for i in range(10):
    displaychar(mu[i])
 
#average shape of numbers
 
"""%%time
# 결과 테스트
k = 10

score = np.zeros(( len(test_labels), k))

for label in range( 0, k ):
    rv = multivariate_normal(mean=mu[label], cov=cov_mat[label])
    for i in range( 0, len( test_labels )):
        score[i, label] = np.log( pi[label] ) + rv.logpdf(test_data[i , :])
predictions = np.argmax(score, axis = 1)        

errors = np.sum( predictions != test_labels )
print("model makes " +str(errors/100000)  + " errors out of 10000")
"""