NARX Time series prediction

1 次查看(过去 30 天)
ADC
ADC 2018-10-17
评论: Greg Heath 2018-10-21
Hi I wish evalueta a Neural network output afther treined (NARX) but I stil can't; Afther trained and salve the net with 2 input and 1 output, I try to type on Command Window the following: "net(A)" where A is a test matrix 1X2 but the answer is: Number of input does not match net.inputs. I tryed whit net(A,0,0) but still the same problem
  1 个评论
Greg Heath
Greg Heath 2018-10-17
Thank you for all of the details we needed to help you.
Greg

请先登录,再进行评论。

回答(1 个)

Greg Heath
Greg Heath 2018-10-18
Don't take Greg too seriously. Sometimes he doesn't take enough time when reading posts. Since
size(A) = [ 1 2 ]
A has to be transposed to be a valid input.
Thank you for formally accepting my answer
Greg
  2 个评论
ADC
ADC 2018-10-20
Hi Greg thanks for your answer but I've alredy tried with the matrix trasposed, and it still does'nt work; Now I noticed that if I train the net with 1000X2 matrix input and the output (target) is 1000x1, I notice that I get no error when I give to the net a new input A=[1x3] , so I mean that if I give to the net the input plus the output the net give me an answer; now I've to work on the ans I got... It does'nt look anything like the target; (I've obtained a costant value, not what I was looking for, even if i change the n of layer ...) where I m wrong??
Greg Heath
Greg Heath 2018-10-21
The input and target sizes are [I N ] = [2 1000] and [O N ] = [1 1000].
It is confusing (aka annoying) when you use the transposes.
More importantly, it could be the source of your trouble.

请先登录,再进行评论。

类别

Help CenterFile Exchange 中查找有关 Sequence and Numeric Feature Data Workflows 的更多信息

Community Treasure Hunt

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

Start Hunting!

Translated by