# -*- coding: utf-8 -*-

 """

 Created on Thu Mar  9 19:20:51 2017

 @author: Jarvis

 """

 '''

 tnesorflow 做机器学习的几个步骤

 1.定义公式

 2.定义loss function，选择优化器，并制定优化器优化loss

 3.迭代地对数据进行训练

 4。在测试集或验证集对准确率进行评测

 '''

 import tensorflow as tf

 import pandas as pd

 import random

 #自己定义的一个选取batch进行训练的一个取batch函数

 def next_batch(mnist, num,ilen = 55):

     size = len(mnist)

     selected_n = set([])

     while(len(selected_n) < num):

         t = random.choice(range(size))

         selected_n.add(t)

     l  = list(selected_n)

     batch_xs = []

     batch_ys = []

     batch_xs = mnist.iloc[l,range(2,54)]

     batch_ys = mnist.iloc[l,range(54,62)]

     return batch_xs,batch_ys

 #对数据进行读取

 org_mnist = pd.read_csv("NDVI_NDWI.csv",header = None,encoding = 'gbk')

 mnist = pd.get_dummies(org_mnist)

 #创建session

 #input_data.read_data_sets("MNIST_data/",one_hot = True)

 sess = tf.InteractiveSession()

 #定义算法公式，在此处就是神经网络的结构方式

 in_units = 52#每一条instance具有52个输入

 h1_units = 30

 h2_units = 20

 h3_units = 10

 h4_units = 5

 #tf.truncated_normal是正态分布的一个东东，主要用于初始化一些W矩阵

 W1 = tf.Variable(tf.truncated_normal([in_units,h1_units],stddev = 0.1))

 b1 = tf.Variable(tf.zeros([h1_units]))

 W2 = tf.Variable(tf.zeros([h1_units,h2_units]))#[h1_units,8]

 b2 = tf.Variable(tf.zeros([h2_units]))#

 W3 = tf.Variable(tf.zeros([h2_units,h3_units]))

 b3 = tf.Variable(tf.zeros([h3_units]))

 W4 = tf.Variable(tf.zeros([h3_units,8]))

 b4 = tf.Variable(tf.zeros([8]))

 '''

 W4 = tf.Variable(tf.zeros([h3_units,h4_units]))

 b4 = tf.Variable(tf.zeros([h4_units]))

 W5 = tf.Variable(tf.zeros([h4_units,8]))

 b5 = tf.Variable(tf.zeros([8]))

 '''

 x = tf.placeholder(tf.float32,[None, in_units])

 keep_prob = tf.placeholder(tf.float32)#dropout 的比例 keep_prob

 hidden1 = tf.nn.sigmoid(tf.matmul(x,W1)+b1)

 hidden1_drop = tf.nn.dropout(hidden1,keep_prob)

 hidden2 = tf.nn.sigmoid(tf.matmul(hidden1_drop,W2)+b2)

 hidden2_drop = tf.nn.dropout(hidden2,keep_prob)

 hidden3 = tf.nn.sigmoid(tf.matmul(hidden2_drop,W3)+b3)

 hidden3_drop = tf.nn.dropout(hidden3,keep_prob)

 #hidden4 = tf.nn.sigmoid(tf.matmul(hidden3_drop,W4)+b4)

 #hidden4_drop = tf.nn.dropout(hidden4,keep_prob)

 y = tf.nn.softmax(tf.matmul(hidden3_drop,W4)+b4)

 y_ = tf.placeholder(tf.float32,[None,8])#[None,10]

 #设置优化函数

 cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_*tf.log(y),reduction_indices=[1]))

 train_step = tf.train.AdagradOptimizer(0.3).minimize(cross_entropy)

 tf.global_variables_initializer().run()

 for i in range(2010):#

     batch_xs, batch_ys = next_batch(mnist,1000)#1000 3

     train_step.run( {x : batch_xs, y_ : batch_ys,keep_prob: 1})

 correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))

 accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

 batch_xs, batch_ys = next_batch(mnist,10000)

 print(accuracy.eval({x:batch_xs,y_:batch_ys,keep_prob:1.0}))