With what parameters we can identify whether the neural network is properly trained or not

1 次查看（过去 30 天）
Nithya SIvasamy
Nithya SIvasamy 2017-9-19
评论： Nithya SIvasamy 2017-9-20
I am new to neural networks.I have created & trained the neural network using nprtool in matlab.I am getting a network,but doesn't know whether the network properly works or whether the network has to be trained again.
With what parameters I can know this.When the neural network will be trained properly.

请先登录，再回答此问题。

采纳的回答

Greg Heath
Greg Heath 2017-9-20
Use NMSE, the normalized-mean-square-error, related to the Rsquare (See Wikipedia) statistic:
0 <= NMSE = 1-Rsquare <= 1
NMSE = mse(target-output)/MSEref
where the reference MSE is obtained from the naïve constant output model
outputc = mean(target,2)
MSEref = mse(target - mean(target,2))
= mean(var(target',1))
If your NMSE is > 1, that means that your design is worse than just assuming the output is a constant!
Hope this helps
Thank you for formally accepting my answer
Greg
PS search both NEWSREADER and ANSWERS using
greg NMSE

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Sequence and Numeric Feature Data Workflows 的更多信息

标签

Community Treasure Hunt

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

Start Hunting!

Translated by