Performance comparison plotting for different Back propagation algorithms

3 次查看（过去 30 天）
Marimuthu Ananthavelu
评论： Marimuthu Ananthavelu 2017-5-9
I am implementing various Backpropagation algorithms for the same dataset and trying to compare the performance. I got a help from the following tutorial for the same.
https://nl.mathworks.com/help/nnet/ug/choose-a-multilayer-neural-network-training-function.html
I tried to plot:
1.mean square error versus execution time for each algorithm 2.time required to converge versus the mean square error convergence goal for each algorithm
Have used the following code, to create my neural network and willing to know how can i implement the above two plots.
[x, t] = bodyfat_dataset;
net = feedforwardnet(10, 'trainlm');
net = train(net, x, t);
y = net(x);

请先登录，再回答此问题。

采纳的回答

Santhana Raj
Santhana Raj 2017-5-4
You have to calculate the mean square error and time taken. You can use (t-y) to calculate the error and tic/toc to calculate the time taken. With this, you can plot a scatter plot for different algorithms.
For the second plot, in the properties of net, you can set the convergence goal. Change it in every iteration and as before use tic & toc to calculate the time taken. With that you can plot your required graph.

更多回答（1 个）

Greg Heath
Greg Heath 2017-5-6
Your stopping criterion for regression should be a fraction of the mean target variance. I tend to choose 0.01 (Rsquare = 0.99) for regression and 0.001 or 0.005 for time-series. For example,
Msegoal = 0.01 * mse(t-y)/ vart1
where
vart1 = mean(var(target',1))
Hope this helps.
Greg

类别

在 Help Center 和 File Exchange 中查找有关 Modeling and Prediction with NARX and Time-Delay Networks 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by