Question

我有一个具有45个属性的数据。首先输入44，输出属性45。所有数据类型均为双精度。在进行深度学习之前，我将所有数据标准化。

options = trainingOptions('sgdm', ...
    'MaxEpochs',400, ...
    'Verbose',true, ...
    'Plots','training-progress');
%trainingDataMale = [XMaleTraining, YMaleTraining];
load swls_deepnetwork_V2;

trainedNetMale = trainNetwork(XMaleTraining, YMaleTraining,swls_deepnetwork_V2,options);

save('trainedNetMale.mat');

给出此介绍并演示如何调用trainnetwork函数，让我描述我的问题。我的深度学习网络设计（在MathWorks中推荐用于序列输入和回归输出的设计）并由我使用matlab的深度学习网络设计器应用程序创建，提供了以下结果。

Training on single CPU.
|========================================================================================|
|  Epoch  |  Iteration  |  Time Elapsed  |  Mini-batch  |  Mini-batch  |  Base Learning  |
|         |             |   (hh:mm:ss)   |     RMSE     |     Loss     |      Rate       |
|========================================================================================|
|       1 |           1 |       00:00:01 |         0.59 |          0.2 |          0.0100 |
|      50 |          50 |       00:00:04 |          NaN |          NaN |          0.0100 |
|     100 |         100 |       00:00:07 |          NaN |          NaN |          0.0100 |
|     150 |         150 |       00:00:10 |          NaN |          NaN |          0.0100 |
|     200 |         200 |       00:00:13 |          NaN |          NaN |          0.0100 |
|     250 |         250 |       00:00:16 |          NaN |          NaN |          0.0100 |
|     300 |         300 |       00:00:19 |          NaN |          NaN |          0.0100 |
|     350 |         350 |       00:00:23 |          NaN |          NaN |          0.0100 |
|     400 |         400 |       00:00:26 |          NaN |          NaN |          0.0100 |
|========================================================================================|

训练时间需要几秒钟。在此之前，我在基于矩阵计算的多层次深度学习网络设计中使用了相同的数据，当时培训需要几天的时间。我在哪里做错了？我尝试了不同的纪元数字和不同的选项。这是我第一次使用matlab深度学习功能，并且我认为自己犯了一些基本错误。任何帮助将不胜感激。提前致谢， FerdaÖzdemirSönmez

深度学习-序列输入回归输出未正确训练

0 个答案: