先尝试建一个单层的神经网络，最终上传kaggle的评估准确率是91.8%.

效果不佳，因为单词神经网络丢失了手写图片的二维结构。

后续将尝试卷积神经网络以提升预测准确率。

import pandas as pd
import numpy as np
import tensorflow as tf

#1 加载数据集，把对输入和结果分开
train = pd.read_csv('DigitRecognizerTrain.csv')
images_train = train.iloc[:,1:].values
labels_train = train.iloc[:,0].values

test = pd.read_csv('DigitRecognizerTest.csv')
images_test = test.iloc[:,:].values

#2 对输入进行处理
images_train = images_train.astype(np.float)
images_train = np.multiply(images_train,1.0/255)
images_test = images_test.astype(np.float)
images_test = np.multiply(images_test,1.0/255)

images_size = images_train.shape[1]

images_width = images_height = np.ceil(np.sqrt(images_size)).astype(np.uint8)

#3 对结果进行处理
labels_count = np.unique(labels_train).shape[0]

#进行one-hot编码
def dense_to_ont_hot(labels_dense,num_classes):
    num_labels = labels_dense.shape[0]
    index_offset = np.arange(num_labels) * num_classes
    labels_one_hot = np.zeros((num_labels,num_classes))
    labels_one_hot.flat[index_offset + labels_dense.ravel()] = 1
    return labels_one_hot

labels = dense_to_ont_hot(labels_train,labels_count)
labels = labels.astype(np.uint8)

#4 对训练集进行分批
batch_size = 64
n_batch = int(len(images_train)/batch_size)

#5 创建一个简单的神经网络用来对图片进行识别
x = tf.placeholder('float',shape=[None,images_size])
y = tf.placeholder('float',shape=[None,labels_count])

weights = tf.Variable(tf.zeros([784,10]))
biases = tf.Variable(tf.zeros([10]))
result = tf.matmul(x, weights) + biases
predictions = tf.nn.softmax(result)

#6 创建损失函数，以交叉熵的平均值为衡量
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=predictions))

#7 用梯度下降法优化参数
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)

#8 初始化变量
init = tf.global_variables_initializer()

#9 计算预测值
with tf.Session() as sess:
    #初始化
    sess.run(init)
    #循环50轮
    for epoch in range(50):
        for batch in range(n_batch-1):
            batch_x = images_train[batch*batch_size:(batch+1)*batch_size] 
            batch_y = labels[batch*batch_size:(batch+1)*batch_size]
            #进行训练
            sess.run(train_step,feed_dict={x:batch_x,y:batch_y})
        batch_x = images_train[n_batch*batch_size:]
        batch_y = labels[n_batch*batch_size:]
        sess.run(train_step,feed_dict={x:batch_x,y:batch_y})
    #计算预测
    myPrediction = sess.run(predictions,feed_dict={x:images_test})

label_test = np.argmax(myPrediction,axis=1)
pd.DataFrame(label_test).to_csv('simpleNN1.csv')