Labfans是一个针对大学生、工程师和科研工作者的技术社区。 | 论坛首页 | 联系我们(Contact Us) |
![]() |
![]() |
#1 |
初级会员
注册日期: 2013-04-11
帖子: 1
声望力: 0 ![]() |
![]()
各位大大好:
小弟目前服務公司,日前給我一個題目做,但由於剛接觸Matlab,對程式操作與理解不足,所以麻煩大大們,幫我解一個ANN題組,以及操作順序,感謝 problem:f(x,y)=sinx+2y² , x,y (1,10) 1.Create a BPNN with one hidden layer of 10 neurons for the problem. General 500 training data, 200 checking data and 100 testing data. Graph your data in 3D mode as in Figure 4.9 on page 114, show your network structure and initial weights in a table as in Table4.4 on page121. 2. Train your network and show the error converge behavior using MSE. Graph your outputs and the targets of both training data and testing data (in separate figures), along with the regression plot and the resulted network weights. 3. Discuss the effect of "number of nodes in hidden layer". What is the best number of nodes in hidden layer? 4. Compare the results from GD, GDM and LM in error converge behavior, R2, training speed, and MAPE. 5. Show your network with the best predictions. |
![]() |
![]() |