机器学习-学习笔记（四） tensorflow+mnist数据集，实现最邻近算法

最新推荐文章于 2020-09-05 01:49:56 发布

Avlon

最新推荐文章于 2020-09-05 01:49:56 发布

阅读量387

点赞数

分类专栏：机器学习

本文链接：https://blog.csdn.net/Avlon/article/details/81296956

版权

机器学习专栏收录该内容

16 篇文章 3 订阅

订阅专栏

KNN在Mnist数据集上的实现，用的是L1距离（各像素差的绝对值的和），tf实现，十分简单的算法在MNIST上效果不错，测试正确率能大概在0.96

import numpy
import input_data
import tensorflow as tf

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

x = tf.placeholder("float", [None, 784])
test = tf.placeholder("float", [784])

distance = tf.reduce_sum(tf.abs(tf.add(x,  tf.negative(test))),axis=1)     

index = tf.arg_min(distance, 0)             
init = tf.global_variables_initializer()

train_sample, train_label = mnist.train.next_batch(10000)
test_sample, test_label = mnist.test.next_batch(500)

sess = tf.Session()
sess.run(init)
accury=0
for i in range(len(test_sample)):
    answer_index = sess.run(index, feed_dict={x: train_sample, test: test_sample[i]})
    print('test:', i, "answer_index:", answer_index, "预测类别:",numpy.argmax(train_label[answer_index]), "真实类别:", numpy.argmax(test_label[i]),)
    if numpy.argmax(train_label[answer_index])==numpy.argmax(test_label[i]):
        accury=accury+1
print(accury/len(test_sample))