tensorflow学习笔记（二）

tensorflow中自带的mnist手写数字识别，运用最简单的单层神经网络，softmax激活函数，极客学院上说准确率有91%，我今天调整到了92%！

import tensorflow as tf
import numpy as np
import math
import tensorflow.examples.tutorials.mnist as mn
class Mnist:
    def __init__(self):
        sess = tf.InteractiveSession()
        self.mnist = mn.input_data.read_data_sets("E:\\Python35\\Lib\\site-packages\\tensorflow\\examples\\tutorials\\mnist\\MNIST_data",one_hot=True)
        self.x = tf.placeholder("float", shape=[None, 784])
        self.y_ = tf.placeholder("float", shape=[None, 10])
        self.W = tf.Variable(tf.zeros([784,10]))
        self.b = tf.Variable(tf.zeros([10]))
        self.y = tf.nn.softmax(tf.matmul(self.x,self.W) + self.b)
        self.cross_entropy = -tf.reduce_sum(self.y_*tf.log(self.y))
        sess.run(tf.global_variables_initializer())

        self.bestModel = None
        self.bestPredict = 0.0
        self.bestIter = 0
        self.bestRate = 0.0
        self.bestSample = 0

        self.iters = [1000,1200,1400]
        self.rates = [0.01,0.02]
        self.samples = [100,150,200]

    def train(self):
        for iter in self.iters:
            for rate in self.rates:
                train_step = tf.train.GradientDescentOptimizer(rate).minimize(self.cross_entropy)
                for sample in self.samples:
                    self.optimizer(iter, rate, sample, train_step)

    def optimizer(self,iter,rate,sample,train_step):
        for i in range(iter):
            batch = self.mnist.train.next_batch(sample)
            model = train_step.run(feed_dict={self.x: batch[0], self.y_: batch[1]})
        correct_prediction = tf.equal(tf.argmax(self.y, 1), tf.argmax(self.y_, 1))
        accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
        predict = accuracy.eval(feed_dict={self.x: self.mnist.test.images, self.y_: self.mnist.test.labels})
        if predict > self.bestPredict:
            self.bestPredict = predict
            self.bestModel = model
            self.bestIter = iter
            self.bestRate = rate
            self.bestSample = sample

    def output(self):
        print("bestRate:",self.bestRate,"bestIter:",self.bestIter,"bestSample:",self.bestSample,"bestPredict:",self.bestPredict)

if __name__ == '__main__':
    mnist = Mnist()
    mnist.train()
    mnist.output()

E:\Python35\python.exe E:/PycharmProjects/test/com/python/machinelearning/tensorflowTest.py
Extracting E:\Python35\Lib\site-packages\tensorflow\examples\tutorials\mnist\MNIST_data\train-images-idx3-ubyte.gz
Extracting E:\Python35\Lib\site-packages\tensorflow\examples\tutorials\mnist\MNIST_data\train-labels-idx1-ubyte.gz
Extracting E:\Python35\Lib\site-packages\tensorflow\examples\tutorials\mnist\MNIST_data\t10k-images-idx3-ubyte.gz
Extracting E:\Python35\Lib\site-packages\tensorflow\examples\tutorials\mnist\MNIST_data\t10k-labels-idx1-ubyte.gz

bestRate: 0.01 bestIter: 1000 bestSample: 100 bestPredict: 0.9193