查看“Jetson Nano系列教程7：TensorFlow入门介绍(二)”的源代码



一、前言

这一篇分享神经网络(NN)的激励函数，损失函数等相关内容，并用TensorFlow构建运行一个神经网络。
二、神经网络(NN)的激励函数，损失函数、梯度下降法
①、激励函数
什么是激励函数？与上一篇提到神经网络逼近或拟合任意连续函数能力之间有什么关系？
从神经元数学描述出发，神经元之间传递的信息，可理解权重值W为信息传递到下一神经元的强度，b表示神经元的阈值或偏置，对来自上一层所有输入信息产生输出信息的变换关系就是激励函数，如下图


[[File:170538j6yzy6etxeltxnle.png]]



[[File:142959pofx7s3i3kz8dlgi.png]]

图中y表示神经元输出，函数f称为激活函数 ( Activation Function )或转移函数 ( Transfer Function ) 若将阈值b看成是神经元的一个输入x0的权重W0，则上面的式子能够简化为：


[[File:143004a8491fo1878z44ja.png]]

设输入向量<b>X</b>=[x0,x1,x2,...,xn],权重和偏值为向量<b style="text-indent: 28px; white-space: pre-wrap;">W</b>


[[File:143007gdpwdylqdpv944qv.png]]



        则神经元的数学模型为y=f(<b>XW</b>)，对于一个二层神经网络(输入层不算一层)，假设输入层神经元的输出为：
O1 = F1( XW1 )
        则第一层(也称隐藏层)的输出为：
O2 = F2 ( F1( XW1 ) W2 )
        则输出层(第二层)的输出为：
O3 = F3( F2 ( F1( XW1 ) W2 ) W3 )
 若激活函数F1~F3都选用线性函数，那么神经网络的输出O3将是输入X的线性函数。因此，若要做高次函数的逼近就应该选用适当的非线性函数作为激活函数。对于对于多达上百层的的深度网络(DNN)也可证明若使用线性函数作为激活函数时，输入输出只是线性变换。<!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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-->
这里分享相关文章讲解为什么神经网络有逼近任意连续函数能力以及证明:
<li><a cjk="" han="" hei="" helvetica="" href="https://zhuanlan.zhihu.com/p/47883643" micro="" neue="" sans="" sans-serif="" sc="" style="font-size: 13px; background-color: rgb(255, 255, 255);" yahei="">深度学习导论 785 lecture1
</li><li><a cjk="" han="" hei="" helvetica="" href="https://zhuanlan.zhihu.com/p/49496192" micro="" neue="" sans="" sans-serif="" sc="" style="font-size: 13px; background-color: rgb(255, 255, 255); font-family: -apple-system, BlinkMacSystemFont, " yahei="">深度学习导论 785 lecture2 上
</li><li><a cjk="" han="" hei="" helvetica="" href="https://zhuanlan.zhihu.com/p/50133198" micro="" neue="" sans="" sans-serif="" sc="" style="font-size: 13px; background-color: rgb(255, 255, 255); font-family: -apple-system, BlinkMacSystemFont, " yahei="">深度学习导论 785 lecture2 下
</li><li>
神经网络拟合能力1
</li><li>
神经网络拟合能力2
</li>
举个例子，信号与系统分析中的傅里叶变换章节，解释了用无穷个不同频率的弦波信号生成方波信号类似逼近过程的原理，
感兴趣读者可点击参阅相关内容
。
常用非线性的激活函数有sigmoid，ReLu，tansh，softsign，ELU等。


（1)、sigmoid函数常用在浅层神经网络，将输入变换到(0,1)区间


[[File:154052uw3zwd0dol4w0r38.png]]


[[File:154051ci3iib93coob3bab.png]]
<!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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-->


（2）、ReLu线性整流函数，将输入变换到Max(0,x)区间

[[File:154052pf6ffc6d69kayypl.png]]


[[File:154051fgxnmfmzm9j1unuu.png]]


  (3)、tanh双曲正切函数，将输入变换到(-1,1)区间

[[File:154052c9ef6obf64fb96bn.png]]


[[File:154052z3sf91u7f2k82fr7.png]]


了解更多激活函数请点击我参考。

tensorflow使用激活函数的用法是：

<syntaxhighlight lang="python">
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers,Sequential
tf.keras.layers.Dense(512,activation=tf.nn.relu)#构建一层由512个神经元，激活函数为relu神经网络
tf.keras.layers.Dense(256,activation=tf.nn.sigmoid)#构建一层由256个神经元，激活函数为sigmoid神经网络
tf.keras.layers.Dense(128,activation=tf.nn.tanh)#构建一层由128个神经元，激活函数为tanh神经网络</syntaxhighlight>

②、损失函数
什么是损失函数？

在机器学习中通常定义指标来表示一个模型是坏的，这个指标称为成本（<b>cost</b>）或损失（<b>loss</b>），然后尽量最小化这个指标。通俗理解是预测结果和样本目标结果相差多少，比如，输入一张牧羊犬样本图片到一个神经网络，若第一次训练后输出一张鱼的预测图片，接下来每次训练输出预测结果分别是鸡、羊等越来越靠近犬这个物种，说明这个神经网络模型权重偏置等参数在往犬这个物种靠近，损失函数逐渐减小。
常用的损失函数有交叉熵(cross-entropy)、平方损失函数、均方误差函数等。
交叉熵常用在<b>分类问题</b>上，产生于信息论里面的信息压缩编码技术，但是它后来演变成为从博弈论到机器学习等其他领域里的重要技术手段，如0-9手写数字分类问题中输出10个值为概率分布，其中最大值表示对应数字(0-9共10个数字)的可能性最大。它的定义如下:


[[File:182105vmhkb82d0r7ugdvp.png.thumb.png]]


y 是我们预测结果（训练结果）, y' 是样本目标结果。比较粗糙的理解是，在多输出神经网络上，交叉熵是用来衡量我们的预测用于描述真相的低效性。


平方损失函数是<b>线性回归问题</b>常用损失函数，一般用在具体数值的预测，例如一年四季某城市用电量预测，房价预测等。


[[File:182110djo5zshe5x3ohx5l.png.thumb.png]]


损失函数详细分析请参考下面相关文章

<li>
常见的损失函数总结
</li><li>
常用损失函数与参数优化
</li><li>
Tensorflow基础知识---损失函数详解
</li><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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-->


tensorflow中用到的损失函数有tf.nn.relu, tf.nn.elu, tf.nn.softmax, tf.nn.softsign等。
 对于分类问题，可以这样定义损失函数，其中softmax_cross_entropy_with_logits函数将训练结果转化成概率分布。<!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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-->

<syntaxhighlight lang="python">
 #y_为样本目标结果，y为样本训练结果
loss = tf.nn.softmax_cross_entropy_with_logits(y, y_)</syntaxhighlight>  对于回归问题，可以这样定义损失函数

<syntaxhighlight lang="python">
loss = tf.reduce_mean(tf.square(y_ - y))</syntaxhighlight><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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--><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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-->
③、梯度下降法
什么是梯度下降？跟神经网络逼近或拟合任意连续函数能力之间有什么关系？
首先回顾下神经网络工作方式，第一个阶段是，神经网络输入样本数据经过前向传递到达输出层得到预测结果；第二阶段是，使用不同计算方式(使用不同的损失函数)计算样本预测结果和样本目标结果的差异并以此作为调整权重偏置的依据，然后回传到上层网络。要想一个网络能拟合任意连续函数，就需要损失函数越小越好，所以需要约定一个方向使每训练完一批次的样本数据得到预测结果和目标结果的差异越来越小，这个使差异越来越小的方向可认为是我们要的梯度下降。梯度是高等数学中的基本概念，对于一元连续函数可理解是2维坐标轴上对函数某点的导数，对于2元函数可理解成物体在一个山峰上滚落速度最快的方向。如下图所示


[[File:144132cf5kmb5ckd5kl72e.png]]



上图中点C往点D的方向移动可以使函数F(x)=x^2的输出变小，定义学习率η表示C点往D点方向移动多大距离，在神经网络中学习率η表示调整权重偏置幅度，学习率η太小会使神经网络训时间增加，太大会使神经网络训练容易震荡，神经网络不容易收敛。


[[File:143406czf13s3m33uo333m.gif]]

上图中可以看到，二元函数中，不止一个波峰波谷，不同的梯度下降法下降的快慢也不一样，神经网络要拟合逼近或拟合任意连续函数那么训练神经网络就需要进行大量运算，还需结合选取不同的梯度下降法。
更多关于梯度下降的资料请点击下面链接：


<li>
梯度下降法的推导
</li><li>
深入浅出--梯度下降法及其实现
</li><li>
梯度下降法种类
</li><!--5f39ae17-8c62-4a45-bc43-b32064c9388a:W3siYmxvY2tJZCI6Ijg5OTctMTU2NTQyOTk3NTE5NSIsImJsb2NrVHlwZSI6InBhcmFncmFwaCIsInN0eWxlcyI6eyJ0ZXh0LWluZGVudCI6MCwiYWxpZ24iOiJsZWZ0IiwiaW5kZW50IjowLCJsaW5lLWhlaWdodCI6MS43NSwiYmFjay1jb2xvciI6IiIsInBhZGRpbmciOiIifSwidHlwZSI6InBhcmFncmFwaCIsInJpY2hUZXh0Ijp7ImRhdGEiOlt7ImNoYXIiOiLmoq8ifSx7ImNoYXIiOiLluqYifSx7ImNoYXIiOiLkuIsifSx7ImNoYXIiOiLpmY0ifSx7ImNoYXIiOiLms5UifSx7ImNoYXIiOiLnmoQifSx7ImNoYXIiOiLmjqgifSx7ImNoYXIiOiLlr7wifV0sImlzUmljaFRleHQiOnRydWUsImtlZXBMaW5lQnJlYWsiOnRydWV9fV0=-->

tensorflow中使用梯度下降，参数0.01表示学习率η，cross_entropy是前面定义的损失函数：

<syntaxhighlight lang="python">
train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy)</syntaxhighlight><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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-->
三、TensorFlow构建执行神经网络
本小结构建一个二层(输入层不算)神经网络并训练MNIST手写数字集，输入层为784个神经元(将28*28二维图像转变成1维的784个像素数据)，隐藏层为1024个神经元，输出有10个神经元。

<syntaxhighlight lang="python">
#导入需要用到的库
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

#下载MNIST手写数字集，将手写数字集28*28图像变成1维的784个数据，
mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)

#定义占位符，MNIST手写数字集有60000张手写数字图片作训练样本，有10000张手写数字图片														
#作测试样本
#张量x的None表示MNIST手写数字集中的图片编号，784表示降维后图片中每一个像素点
#张量y的None表示MNIST手写数字集中的图片编号，10表示0~9共10个数字概率，y为样本的目					
#标结果(也称标签，导入手写数字集已经捆绑了对应的数字)
x = tf.placeholder(tf.float32, [None,784])
y = tf.placeholder(tf.float32, [None,10])

#定义权重张量w1，偏置张量b1，784个输入层神经元连接1024个隐藏层神经元
#定义隐藏层输出张量a，使用tf.nn.relu激活函数，tf.matmul(x,w1)+b1表示神经元的连接
w1 = tf.Variable(tf.truncated_normal([784,1024]), dtype = tf.float32)
b1 = tf.Variable(tf.zeros([1,1024]), dtype = tf.float32)
a = tf.nn.relu(tf.matmul(x,w1) + b1) 

#定义权重张量w2，偏置张量b2，1024个隐藏层神经元连接10个输出层神经元
#定义输出层张量y_，使用tf.nn.softmax激活函数，tf.matmul(a,w2)+b2表示神经元的连接
w2 = tf.Variable(tf.ones([1024,10]))
b2 = tf.Variable(tf.zeros([1,10]))
y_= tf.nn.softmax(tf.matmul(a,w2) + b2) 

#定义损失函数和梯度下降法
loss = tf.reduce_mean(-tf.reduce_sum(y*tf.log(y_), axis = 1)) 
train_step = tf.train.AdamOptimizer(0.0001).minimize(loss)

#初始化特殊张量
init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)

#每训练一个批次样本数据后后获得识别正确率
correct_prediction = tf.equal(tf.argmax(y_, axis = 1),tf.argmax(y, axis = 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

#定义好上面所有的运行方式后，使用sess.run启动tensorflow
#在1000轮训练中，mnist.train.next_batch(100)表示每轮训练取出100张样本图片，用feed_dict	
#方法把100张样本图片扔进占位符x，y进行训练，train_step参数是上面定义好的网络权重和偏	
#置的调整方法
#每训练100轮就检验一次MNIST手写数据集中的10000张测试图片
for i in range(1000):
    batch_xs, batch_ys = mnist.train.next_batch(100)
    sess.run(train_step, feed_dict = {x:batch_xs,y:batch_ys})
    if i % 100 == 0:
        print(sess.run(accuracy, feed_dict={x: mnist.test.images, y: mnist.test.labels}))</syntaxhighlight>
实际训练结果如下图所示


[[File:170908wpqzwqq24qow2qey.png]]

第一次训练时，输出准确率只有37.16%，最终训练的准确率为94.27%。感兴趣读者可参考上面代码在JetsonNano上进行测试。
MNIST手写数字集相关请参考下面链接：

<li>
利用神经网络识别手写数字
</li>

<!--5f39ae17-8c62-4a45-bc43-b32064c9388a:W3siYmxvY2tJZCI6IjcxODItMTU2NTg1MDA0ODk5MyIsImJsb2NrVHlwZSI6InBhcmFncmFwaCIsInN0eWxlcyI6eyJhbGlnbiI6ImxlZnQiLCJpbmRlbnQiOjAsInRleHQtaW5kZW50IjowLCJsaW5lLWhlaWdodCI6MS43NSwiYmFjay1jb2xvciI6IiIsInBhZGRpbmciOiIifSwidHlwZSI6InBhcmFncmFwaCIsInJpY2hUZXh0Ijp7ImRhdGEiOlt7ImNoYXIiOiLliKkiLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6IueUqCIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi56WeIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLnu48iLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6Iue9kSIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi57ucIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLor4YiLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6IuWIqyIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi5omLIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLlhpkiLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6IuaVsCIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi5a2XIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fV0sImlzUmljaFRleHQiOnRydWUsImtlZXBMaW5lQnJlYWsiOnRydWV9fV0=-->
以上资料由waveshare team整理。





<!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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--><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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--><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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--><!--5f39ae17-8c62-4a45-bc43-b32064c9388a:W3siYmxvY2tJZCI6IjE1NDUtMTU2NTc0OTIyMzMzMiIsImJsb2NrVHlwZSI6InBhcmFncmFwaCIsInN0eWxlcyI6eyJ0ZXh0LWluZGVudCI6MSwiYWxpZ24iOiJsZWZ0IiwiaW5kZW50IjowLCJsaW5lLWhlaWdodCI6MS4yLCJiYWNrLWNvbG9yIjoiIiwicGFkZGluZyI6IiJ9LCJ0eXBlIjoicGFyYWdyYXBoIiwicmljaFRleHQiOnsiZGF0YSI6W3siY2hhciI6IlIifSx7ImNoYXIiOiJlIn0seyJjaGFyIjoiTCJ9LHsiY2hhciI6InUifSx7ImNoYXIiOiLvvIgifSx7ImNoYXIiOiLvvIkifSx7ImNoYXIiOiLnur8ifSx7ImNoYXIiOiLmgKcifSx7ImNoYXIiOiLmlbQifSx7ImNoYXIiOiLmtYEifSx7ImNoYXIiOiLlh70ifSx7ImNoYXIiOiLmlbAifV0sImlzUmljaFRleHQiOnRydWUsImtlZXBMaW5lQnJlYWsiOnRydWV9fV0=--><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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--><!--5f39ae17-8c62-4a45-bc43-b32064c9388a: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--><!--5f39ae17-8c62-4a45-bc43-b32064c9388a:W3siYmxvY2tJZCI6IjkwNDAtMTU2NTQwNjk3OTQ1NiIsImJsb2NrVHlwZSI6ImhlYWRpbmciLCJzdHlsZXMiOnsiYWxpZ24iOiJsZWZ0IiwiaW5kZW50IjowLCJ0ZXh0LWluZGVudCI6MCwibGluZS1oZWlnaHQiOjEuNzUsImJhY2stY29sb3IiOiIiLCJwYWRkaW5nIjoiIn0sInR5cGUiOiJoZWFkaW5nIiwicmljaFRleHQiOnsiZGF0YSI6W3siY2hhciI6IuS4gCIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoi44CBIiwic3R5bGVzIjp7ImJvbGQiOnRydWUsImZvbnQtc2l6ZSI6MTgsImNvbG9yIjoiIzU5NTk1OSJ9fSx7ImNoYXIiOiLliY0iLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZSwiZm9udC1zaXplIjoxOCwiY29sb3IiOiIjNTk1OTU5In19LHsiY2hhciI6IuiogCIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX1dLCJpc1JpY2hUZXh0Ijp0cnVlLCJrZWVwTGluZUJyZWFrIjp0cnVlfSwibGV2ZWwiOiJiIn0seyJibG9ja0lkIjoiMzg5Mi0xNTY1NDA3MDQwMDg1IiwiYmxvY2tUeXBlIjoicGFyYWdyYXBoIiwic3R5bGVzIjp7InRleHQtaW5kZW50IjoxLCJhbGlnbiI6ImxlZnQiLCJpbmRlbnQiOjAsImxpbmUtaGVpZ2h0IjoxLjc1LCJiYWNrLWNvbG9yIjoiIiwicGFkZGluZyI6IiJ9LCJ0eXBlIjoicGFyYWdyYXBoIiwicmljaFRleHQiOnsiZGF0YSI6W3siY2hhciI6Iui/mSJ9LHsiY2hhciI6IuS4gCJ9LHsiY2hhciI6IuevhyJ9LHsiY2hhciI6IuWIhiJ9LHsiY2hhciI6IuS6qyJ9LHsiY2hhciI6IuelniJ9LHsiY2hhciI6Iue7jyJ9LHsiY2hhciI6Iue9kSJ9LHsiY2hhciI6Iue7nCJ9LHsiY2hhciI6IigifSx7ImNoYXIiOiJOIn0seyJjaGFyIjoiTiJ9LHsiY2hhciI6IikifSx7ImNoYXIiOiLnmoQifSx7ImNoYXIiOiLmv4AiLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZX19LHsiY2hhciI6IuWKsSIsInN0eWxlcyI6eyJib2xkIjp0cnVlfX0seyJjaGFyIjoi5Ye9Iiwic3R5bGVzIjp7ImJvbGQiOnRydWV9fSx7ImNoYXIiOiLmlbAiLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZX19LHsiY2hhciI6Iu+8jCIsInN0eWxlcyI6eyJib2xkIjp0cnVlfX0seyJjaGFyIjoi5o2fIiwic3R5bGVzIjp7ImJvbGQiOnRydWV9fSx7ImNoYXIiOiLlpLEiLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZX19LHsiY2hhciI6IuWHvSIsInN0eWxlcyI6eyJib2xkIjp0cnVlfX0seyJjaGFyIjoi5pWwIiwic3R5bGVzIjp7ImJvbGQiOnRydWV9fSx7ImNoYXIiOiLnm7gifSx7ImNoYXIiOiLlhbMifSx7ImNoYXIiOiLlhoUifSx7ImNoYXIiOiLlrrkifSx7ImNoYXIiOiLvvIwifSx7ImNoYXIiOiLlubYifSx7ImNoYXIiOiLnlKgifSx7ImNoYXIiOiJUIn0seyJjaGFyIjoiZSJ9LHsiY2hhciI6Im4ifSx7ImNoYXIiOiJzIn0seyJjaGFyIjoibyJ9LHsiY2hhciI6InIifSx7ImNoYXIiOiJGIn0seyJjaGFyIjoibCJ9LHsiY2hhciI6Im8ifSx7ImNoYXIiOiJ3In0seyJjaGFyIjoi5p6EIn0seyJjaGFyIjoi5bu6In0seyJjaGFyIjoi6L+QIn0seyJjaGFyIjoi6KGMIn0seyJjaGFyIjoi5LiAIn0seyJjaGFyIjoi5LiqIn0seyJjaGFyIjoi56WeIn0seyJjaGFyIjoi57uPIn0seyJjaGFyIjoi572RIn0seyJjaGFyIjoi57ucIn0seyJjaGFyIjoi44CCIn1dLCJpc1JpY2hUZXh0Ijp0cnVlLCJrZWVwTGluZUJyZWFrIjp0cnVlfX0seyJibG9ja0lkIjoiMTc0OC0xNTY1NDA3MjEyMzg1IiwiYmxvY2tUeXBlIjoiaGVhZGluZyIsInN0eWxlcyI6eyJhbGlnbiI6ImxlZnQiLCJpbmRlbnQiOjAsInRleHQtaW5kZW50IjowLCJsaW5lLWhlaWdodCI6MS43NSwiYmFjay1jb2xvciI6IiIsInBhZGRpbmciOiIifSwidHlwZSI6ImhlYWRpbmciLCJyaWNoVGV4dCI6eyJkYXRhIjpbeyJjaGFyIjoi5LqMIiwic3R5bGVzIjp7ImJvbGQiOnRydWUsImZvbnQtc2l6ZSI6MTgsImNvbG9yIjoiIzU5NTk1OSJ9fSx7ImNoYXIiOiLjgIEiLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZSwiZm9udC1zaXplIjoxOCwiY29sb3IiOiIjNTk1OTU5In19LHsiY2hhciI6IuelniIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoi57uPIiwic3R5bGVzIjp7ImJvbGQiOnRydWUsImZvbnQtc2l6ZSI6MTgsImNvbG9yIjoiIzU5NTk1OSJ9fSx7ImNoYXIiOiLnvZEiLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZSwiZm9udC1zaXplIjoxOCwiY29sb3IiOiIjNTk1OTU5In19LHsiY2hhciI6Iue7nCIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoiKCIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoiTiIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoiTiIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoiKSIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoi55qEIiwic3R5bGVzIjp7ImJvbGQiOnRydWUsImZvbnQtc2l6ZSI6MTgsImNvbG9yIjoiIzU5NTk1OSJ9fSx7ImNoYXIiOiLmv4AiLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZSwiZm9udC1zaXplIjoxOCwiY29sb3IiOiIjNTk1OTU5In19LHsiY2hhciI6IuWKsSIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoi5Ye9Iiwic3R5bGVzIjp7ImJvbGQiOnRydWUsImZvbnQtc2l6ZSI6MTgsImNvbG9yIjoiIzU5NTk1OSJ9fSx7ImNoYXIiOiLmlbAiLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZSwiZm9udC1zaXplIjoxOCwiY29sb3IiOiIjNTk1OTU5In19LHsiY2hhciI6Iu+8jCIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoi5o2fIiwic3R5bGVzIjp7ImJvbGQiOnRydWUsImZvbnQtc2l6ZSI6MTgsImNvbG9yIjoiIzU5NTk1OSJ9fSx7ImNoYXIiOiLlpLEiLCJzdHlsZXMiOnsiYm9sZCI6dHJ1ZSwiZm9udC1zaXplIjoxOCwiY29sb3IiOiIjNTk1OTU5In19LHsiY2hhciI6IuWHvSIsInN0eWxlcyI6eyJib2xkIjp0cnVlLCJmb250LXNpemUiOjE4LCJjb2xvciI6IiM1OTU5NTkifX0seyJjaGFyIjoi5pWwIiwic3R5bGVzIjp7ImJvbGQiOnRydWUsImZvbnQtc2l6ZSI6MTgsImNvbG9yIjoiIzU5NTk1OSJ9fV0sImlzUmljaFRleHQiOnRydWUsImtlZXBMaW5lQnJlYWsiOnRydWV9LCJsZXZlbCI6ImIifSx7ImJsb2NrSWQiOiI2Njc0LTE1NjU2ODk0NzYxMzUiLCJibG9ja1R5cGUiOiJwYXJhZ3JhcGgiLCJzdHlsZXMiOnsidGV4dC1pbmRlbnQiOjEsImFsaWduIjoibGVmdCIsImluZGVudCI6MCwibGluZS1oZWlnaHQiOjEuNzUsImJhY2stY29sb3IiOiIiLCJwYWRkaW5nIjoiIn0sInR5cGUiOiJwYXJhZ3JhcGgiLCJyaWNoVGV4dCI6eyJkYXRhIjpbeyJjaGFyIjoi4pGgIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLjgIEiLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6Iua/gCIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi5YqxIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLlh70iLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6IuaVsCIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX1dLCJpc1JpY2hUZXh0Ijp0cnVlLCJrZWVwTGluZUJyZWFrIjp0cnVlfX0seyJibG9ja0lkIjoiNDI0My0xNTY1NzQ5MjUxNDYwIiwiYmxvY2tUeXBlIjoicGFyYWdyYXBoIiwic3R5bGVzIjp7InRleHQtaW5kZW50IjoxLCJhbGlnbiI6ImxlZnQiLCJpbmRlbnQiOjAsImxpbmUtaGVpZ2h0IjoxLjc1LCJiYWNrLWNvbG9yIjoiIiwicGFkZGluZyI6IiJ9LCJ0eXBlIjoicGFyYWdyYXBoIiwicmljaFRleHQiOnsiZGF0YSI6W3siY2hhciI6IuS7gCIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi5LmIIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLmmK8iLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6Iua/gCIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi5YqxIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLlh70iLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6IuaVsCIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi77yfIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLkuI4iLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkzOTM5In19LHsiY2hhciI6IuS4iiIsInN0eWxlcyI6eyJjb2xvciI6IiMzOTM5MzkifX0seyJjaGFyIjoi5LiAIiwic3R5bGVzIjp7ImNvbG9yIjoiIzM5MzkzOSJ9fSx7ImNoYXIiOiLnr4ciLCJzdHlsZXMiOnsiY29sb3IiOiIjMzkz-->







 

<b>Tab标签:</b> 
Jetson Nano系列教程 TensorFlow python