# coding: utf-8

# # Laborator 3

# In[1]:


import numpy as np
import matplotlib.pyplot as plt


# In[2]:

#change this path
dataPath = "C:/Users/Student/Desktop/17lab3/data/data/"

#load train images
train_images = np.loadtxt(dataPath + "train_images.txt")
train_labels = np.loadtxt(dataPath + "train_labels.txt",'int8')
print(train_images.shape)
print(train_images.ndim)
print(type(train_images[0,0]))
print(train_images.size)
print(train_images.nbytes)


# In[3]:


#plot the first 100 training images with their labels in a 10 x 10 subplot
nbImages = 10
plt.figure(figsize=(5,5))
for i in range(nbImages**2):
    plt.subplot(nbImages,nbImages,i+1)
    plt.axis('off')
    plt.imshow(np.reshape(train_images[i,:],(28,28)),cmap = "gray")
#plt.show()
labels_nbImages = train_labels[:nbImages**2]
#print(np.reshape(labels_nbImages,(nbImages,nbImages)))


# In[4]:


#load test images
test_images = np.loadtxt(dataPath + "test_images.txt")
test_labels = np.loadtxt(dataPath + "test_labels.txt",'int8')
#plot the first 100 testing images with their labels in a 10 x 10 subplot
nbImages = 10
plt.figure(figsize=(5,5))
for i in range(nbImages**2):
    plt.subplot(nbImages,nbImages,i+1)
    plt.axis('off')
    plt.imshow(np.reshape(test_images[i,:],(28,28)),cmap = "gray")
#plt.show()
labels_nbImages = test_labels[:nbImages**2]
#print(np.reshape(labels_nbImages,(nbImages,nbImages)))


# In[ ]:


#do 1-NN, 3-NN, 5-NN, 7 -NN for the first test image
#plot the neighbors

v=test_images[0,:]
distanta=np.sqrt(np.sum((v-train_images)**2,axis=1))
indici=distanta.argsort();
print("Predictia mea este", train_labels[indici[0]])

#16:00

def clasifica( elem, train, labels, k=1, dist='l1'):

    if dist == 'l2':
        distanta = np.sqrt(np.sum((elem-train)**2,axis=1))
    elif dist =='l1':
        distanta = np.sum(np.abs(elem-train),axis=1)
    else:
        print('eroare')
        #return -1

    indici = distanta.argsort()
    vecini=labels[indici[0:k]]
    contor=np.bincount(vecini)

    return contor.argmax()

def acuratete(p,y):
    p = np.array(p)
    y = np.array(y)
    return len(p[p==y])*100/len(y)

l,c = test_images.shape

for k in [1,3,5,7]:
    predictii=[]
    for i in range (0,l):
        elem = test_images[i,:]
        p = clasifica(elem,train_images,train_labels,k)
        predictii.append(p)
    print(acuratete(predictii,test_labels))