win7 x64 vs2015 c++ 使用 tensorflow - 1 简单神经网络训练,预测

最新推荐文章于 2021-04-14 15:41:44 发布

zsyddl2

最新推荐文章于 2021-04-14 15:41:44 发布

阅读量1.2k

点赞数

分类专栏：机器人、人工智能深度学习 tensorflow c c++ 文章标签： tensorflow vs2015 c++

本文链接：https://blog.csdn.net/zsyddl2/article/details/80973388

版权

c++ 同时被 3 个专栏收录

20 篇文章

订阅专栏

13 篇文章

订阅专栏

机器人、人工智能

11 篇文章

订阅专栏

这是一个有一个中间层一个输出层的神经网络demo.

输入数据为 1维数据.中间层10个神经元输出成10个神经元. 其中一些都是临时为了测试修改的, 数据也是写固定了.

后面补充可以动态输入数据的情况. 完整工程网盘路径: 附带了 tensorflow lib 所以很大.

链接：https://pan.baidu.com/s/1fwP7-D1-Usz1Ec5cY-RaYg 密码：0hlb

#include <fstream>
#include <utility>
#include <vector>
#include <Eigen/Core>
#include <Eigen/Dense>

#include "tensorflow/cc/ops/const_op.h"
#include "tensorflow/cc/ops/image_ops.h"
#include "tensorflow/cc/ops/standard_ops.h"
#include "tensorflow/core/framework/graph.pb.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/graph/default_device.h"
#include "tensorflow/core/graph/graph_def_builder.h"
#include "tensorflow/core/lib/core/errors.h"
#include "tensorflow/core/lib/core/stringpiece.h"
#include "tensorflow/core/lib/core/threadpool.h"
#include "tensorflow/core/lib/io/path.h"
#include "tensorflow/core/lib/strings/stringprintf.h"
#include "tensorflow/core/platform/env.h"
#include "tensorflow/core/platform/init_main.h"
#include "tensorflow/core/platform/logging.h"
#include "tensorflow/core/platform/types.h"
#include "tensorflow/core/public/session.h"
#include "tensorflow/core/util/command_line_flags.h"

#include "tensorflow/cc/client/client_session.h"
#include "tensorflow/cc/framework/gradients.h"
using namespace std;
using namespace tensorflow;
using namespace tensorflow::ops;
using tensorflow::Flag;
using tensorflow::Tensor;
using tensorflow::Status;
using tensorflow::string;
using tensorflow::int32;

void TestTrain10Value()
{
   Scope root = Scope::NewRootScope();

   //定义placeholder
   Placeholder phX = Placeholder(root, DT_DOUBLE);
   Placeholder phY = Placeholder(root, DT_DOUBLE);

   //定义神经网络中间层
   //一列代表一个神经元。列表神经元个数
   //定义10个神经元,每个神经元1维数据.
   auto varWL1 = Variable(root, { 1, 10 }, DT_DOUBLE);
   auto Weights_L1 = Assign(root, varWL1, { { 0.001, 0.002, 0.012, 0.003, 0.004, 0.001, 0.002, 0.0015, 0.0018, 0.007 } });
   //b 的列必须和的列对上。也就是10个神经元需要10个b
   auto varBL1 = Variable(root, { 1, 10 }, DT_DOUBLE);
   //    auto biases_L1 = Assign(root, vBL1, ZerosLike(root, { 1, 10 }));
   auto biases_L1 = Assign(root, varBL1, { { 0.010, 0.011, 0.012, 0.009, 0.002, 0.003, 0.001, 0.007, 0.005, 0.009 } });
   auto matmul = MatMul(root, phX, varWL1);
   auto Wx_plus_b_L1 = Add(root, matmul, varBL1);
//   auto Wx_plus_b_L1 = Add(root, MatMul(root, phX, varWL1), varBL1);
   //   auto Wx_plus_b_L1 = Add(root, Mul(root, phX, varWL1), varBL1);
   auto L1 = Sin(root, Wx_plus_b_L1);
   //layer2, 10个神经元每个神经元 W维度为10
   auto varWL2 = Variable(root, { 10, 10 }, DT_DOUBLE);
   //auto Weights_L1 = Assign(root, varWL1, { { 0.01, 0.02}, {0.01, 0.02 }});
   auto Weights_L2 = Assign(root, varWL2, { { 0.001, 0.002, 0.012, 0.003, 0.004, 0.001, 0.002, 0.0015, 0.0018, 0.007 },
   { 0.001, 0.002, 0.012, 0.003, 0.004, 0.001, 0.002, 0.0015, 0.0018, 0.007 } ,
   { 0.001, -0.002, 0.012, 0.003, 0.004, 0.001, 0.002, 0.0015, 0.0018, 0.007 } ,
   { 0.001, 0.002, -0.012, 0.003, 0.004, 0.001, 0.002, 0.0015, 0.0018, 0.007 },
   { 0.001, 0.002, 0.012, -0.003, 0.004, 0.001, 0.002, 0.0015, 0.0018, 0.007 },
   { 0.001, 0.002, 0.012, 0.003,-0.004, 0.001, 0.002, 0.0015, 0.0018, 0.007 },
   { 0.001, 0.002, 0.012, 0.003, 0.004, -0.001, 0.002, 0.0015, 0.0018, 0.007 },
   { 0.001, 0.002, 0.012, 0.003, 0.004, 0.001, -0.002, 0.0015, 0.0018, 0.007 },
   { 0.001, 0.002, 0.012, 0.003, 0.004, 0.001, 0.002, -0.0015, 0.0018, 0.007 },
   { 0.001, 0.002, 0.012, 0.003, 0.004, 0.001, 0.002, 0.0015, -0.0018, 0.007 }
   });
   //b 的列必须和 w的列对上。
   auto varBL2 = Variable(root, { 1, 10 }, DT_DOUBLE);
   //    auto biases_L1 = Assign(root, vBL1, ZerosLike(root, { 1, 10 }));
   auto biases_L2 = Assign(root, varBL2, { { 0.010, 0.011, 0.012, 0.009, 0.002, 0.003, 0.001, 0.007, 0.005, 0.009 } });
   auto Wx_plus_b_L12 = Add(root, MatMul(root, L1, varWL2), varBL2);
   //   auto Wx_plus_b_L1 = Add(root, Mul(root, phX, varWL1), varBL1);
   auto L2 = Sin(root, Wx_plus_b_L12);
   // 定义神经网络输出层
   //
   auto prediction = L2;
   //
   //损失函数
   auto square = Square(root, Subtract(root, prediction, phY));
   //auto loss = Subtract(root, phY, prediction );//ReduceMean(root, square, { 0, 1 });

   auto loss = square;// ReduceMean(root, square, { 0, 1 });

   std::vector<Output> grad_outputs;
   TF_CHECK_OK(AddSymbolicGradients(root, { loss }, { varWL1, varBL1, varWL2, varBL2 }, &grad_outputs));

   // update the weights and bias using gradient descent
   auto apply_w1 = ApplyGradientDescent(root, varWL1, Cast(root, 0.03, DT_DOUBLE), { grad_outputs[0] });
   auto apply_b1 = ApplyGradientDescent(root, varBL1, Cast(root, 0.03, DT_DOUBLE), { grad_outputs[1] });
   auto apply_w2 = ApplyGradientDescent(root, varWL2, Cast(root, 0.03, DT_DOUBLE), { grad_outputs[2] });
   auto apply_b2 = ApplyGradientDescent(root, varBL2, Cast(root, 0.03, DT_DOUBLE), { grad_outputs[3] });
   // 使用梯度下降法训练

   //auto train_step = train.GradientDescentOptimizer(0.3).minimize(loss);
   //   auto train_step = ApplyGradientDescent(root, loss, { 1 }, { 0.3 });
   //x= 0.3, y=0.6
   std::vector<Tensor> outputs;
   ClientSession session(root);
   //TF_CHECK_OK(session.Run({ Weights_L1, biases_L1 }, &outputs));
   TF_CHECK_OK(session.Run({ Weights_L1, biases_L1, Weights_L2, biases_L2 }, &outputs));
   //    std::cout << outputs[0].matrix<double>() << std::endl << outputs[1].matrix<double>() << std::endl;
   //    //{{ 5.0 },{4.0}}代表2个1维数据， {{ 5.0 ,4.0}}代表1个2维数据
   //    TF_CHECK_OK(session.Run({ { phX,{ { 5.0 },{4.0} } } }, { Wx_plus_b_L1 }, {}, &outputs));
   //    std::cout << outputs[0].matrix<double>() << std::endl;

   for (int i = 0; i < 20000; ++i) {

       TF_CHECK_OK(session.Run({ { phX,{ { 1.0 },{ 2.0 },{ 3.0 },{ 4.0 },{ 5.0 },{ 6.0 },{ 7.0 },{ 8.0 } ,{ 9.0 },{ 10.0 } } },
       { phY,{ { 1.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },{ 0.0 , 1.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },
       { 0.0 , 0.0, 1.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },{ 0.0 , 0.0, 0.0 , 1.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },
       { 0.0 , 0.0, 0.0 , 0.0, 1.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },{ 0.0 , 0.0, 0.0 , 0.0, 0.0 , 1.0, 0.0 , 0.0, 0.0 , 0.0 },
       { 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 1.0 , 0.0, 0.0 , 0.0 },{ 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 1.0, 0.0 , 0.0 },
       { 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 1.0 , 0.0 },{ 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 1.0 } } }
       },
       { apply_w1, apply_b1, apply_w2, apply_b2 },
           &outputs));

       if (i % 100 == 0) {
           TF_CHECK_OK(session.Run({ { phX,{ { 1.0 },{ 2.0 },{ 3.0 },{ 4.0 },{ 5.0 },{ 6.0 },{ 7.0 },{ 8.0 } ,{ 9.0 },{ 10.0 } } },
           { phY,{
               { 1.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },{ 0.0 , 1.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },
               { 0.0 , 0.0, 1.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },{ 0.0 , 0.0, 0.0 , 1.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },
               { 0.0 , 0.0, 0.0 , 0.0, 1.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0 },{ 0.0 , 0.0, 0.0 , 0.0, 0.0 , 1.0, 0.0 , 0.0, 0.0 , 0.0 },
               { 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 1.0 , 0.0, 0.0 , 0.0 },{ 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 1.0, 0.0 , 0.0 },
               { 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 1.0 , 0.0 },{ 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 0.0, 0.0 , 1.0 } } }
           },
           { loss },
               &outputs));
           std::cout << "Loss after " << i << " steps " << outputs[0].matrix<double>() << std::endl;
       }
       //std::cout << outputs[0].matrix<double>() << " " << outputs[1].matrix<double>() << std::endl;
   }

   TF_CHECK_OK(session.Run({ { phX,{ { 1.8 } } } }, { prediction }, &outputs));
   //    cout << "DNN output: " << *outputs[0].matrix<double>().data() << endl;
   cout << "DNN output: " << outputs[0].matrix<double>() << std::endl;
   TF_CHECK_OK(session.Run({ { phX,{ { 2.0 } } } }, { prediction }, &outputs));
   //    cout << "DNN output: " << *outputs[0].matrix<double>().data() << endl;
   cout << "DNN output: " << outputs[0].matrix<double>() << std::endl;
   TF_CHECK_OK(session.Run({ { phX,{ { 3.0 } } } }, { prediction }, &outputs));
   //    cout << "DNN output: " << *outputs[0].matrix<double>().data() << endl;
   cout << "DNN output: " << outputs[0].matrix<double>() << std::endl;
}