以下代碼數(shù)據(jù)源是mnist聂儒，運(yùn)用softmax做分類器灼擂，使用交叉熵做損失函數(shù)。
這個(gè)腳本的目的不是保證分類器的效果焰盗，只供學(xué)習(xí)tensorflow編程使用璧尸，后續(xù)加入tensorflow的制圖功能
import os
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
import numpy as np
# mnist數(shù)據(jù)集的目錄，先獲得整個(gè)工程中數(shù)據(jù)目錄的根目錄熬拒，然后拼接子目錄
dataPathRoot = os.path.abspath('./../../../data')
mnistDataPath = os.path.join(dataPathRoot, 'MNIST_data')
print(mnistDataPath)
# get the mnist train data
mnist = input_data.read_data_sets(mnistDataPath, one_hot=True)
batch_size=100
# 訓(xùn)練或測(cè)試數(shù)據(jù)輸入
with tf.name_scope('input'):
x = tf.placeholder(tf.float32, [None, 28 * 28], name='input_x')
y = tf.placeholder(tf.float32, [None, 10], name='target_y')
# 需要優(yōu)化的變量
# with tf.name_scope('linefuanction'):
w = tf.Variable(tf.zeros([28 * 28, 10]), name='weight')
b = tf.Variable(tf.zeros(10), name='bias')
pred = tf.nn.softmax(tf.matmul(x, w) + b, name='softmax')
# 交叉熵和梯度下降優(yōu)化
cross_entropy=-tf.reduce_sum(y*tf.log(pred))
train_step=tf.train.GradientDescentOptimizer(0.01, name='GradientDescent').minimize(cross_entropy)
# 模型評(píng)估
correct_prediction=tf.equal(tf.argmax(pred,1),tf.argmax(y,1))
accuray=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
init_variable = tf.initialize_all_variables()
with tf.Session() as sess:
sess.run(init_variable)
# 訓(xùn)練
# for i in range(1000):
# one_batch_x, one_batch_y = mnist.train.next_batch(batch_size)
# sess.run(train_step, feed_dict={x: one_batch_x.astype(np.float32), y: one_batch_y.astype(np.float32)})
one_batch_x, one_batch_y = mnist.train.next_batch(batch_size)
sess.run(train_step, feed_dict={x: one_batch_x, y: one_batch_y})
# 測(cè)試
print(sess.run(accuray,feed_dict={x:mnist.test.images[:batch_size],y:mnist.test.labels[:batch_size]}))
