#!/usr/bin/env python
# coding: utf-8

# In[1]:


import numpy as np


# In[2]:


y = [0.1, 0.05, 0.6, 0.0, 0.05, 0.1, 0.0, 0.1, 0.0, 0.0]


# In[3]:


t = [0, 0, 1, 0, 0, 0, 0, 0, 0, 0]


# In[4]:


def mean_square_error(y, t):
    return 0.5 * np.sum((y - t) ** 2)


# In[6]:


mean_square_error(np.array(y), np.array(t))


# In[7]:


y2 = [0.1, 0.05, 0.1, 0.0, 0.05, 0.1, 0.0, 0.6, 0.0, 0.0]


# In[8]:


mean_square_error(np.array(y2), np.array(t))


# In[9]:


def cross_entropy_error(y, t):
    delta = 1e-7
    return -np.sum(t * np.log(y + delta))


# In[11]:


cross_entropy_error(np.array(y), np.array(t))


# In[12]:


cross_entropy_error(np.array(y2), np.array(t))


# In[13]:


from dataset.mnist import load_mnist


# In[14]:


(x_train, t_train), (x_test, t_test) = \
    load_mnist(normalize=True, one_hot_label=True)


# In[15]:


x_train.shape


# In[16]:


t_train.shape


# In[21]:


train_size = x_train.shape[0]


# In[17]:


batch_size = 10


# In[22]:


batch_mask = np.random.choice(train_size, batch_size)


# In[23]:


x_batch = x_train[batch_mask]
t_batch = t_train[batch_mask]


# In[24]:


x_batch


# In[25]:


t_batch


# In[34]:


def cross_entropy_error(y, t):
    if y.ndim == 1:
        t = t.reshape(1, t.size)
        y = y.reshape(1, y.size)

    delta = 1e-7
    batch_size = y.shape[0]
    return -np.sum(t * np.log(y + delta)) / batch_size


# In[35]:


cross_entropy_error(np.array(y), np.array(t))


# In[33]:


y


# In[36]:


def numerical_diff(f, x):
    h = 1e-4
    return (f(x+h) - f(x-h)) / (2*h)


# In[37]:


def function_1(x):
    return 0.01*x**2 + 0.1*x


# In[38]:


numerical_diff(function_1, 5)


# In[39]:


numerical_diff(function_1, 10)


# In[40]:


def function_2(x0, x1):
    return x0**2 + x1**2


# In[41]:


def function_tmp1(x0):
    return function_2(x0, 4.0)


# In[43]:


numerical_diff(function_tmp1, 3.0)


# In[42]:


def function_tmp2(x1):
    return function_2(3.0, x1)


# In[44]:


numerical_diff(function_tmp2, 4.0)


# In[ ]: