I have 100 matrices each of dimension [200*8000] which forms my final input matrix of dimension [200*800000]. My Target(Output) matrix is of dimension[1*100] i.e. for each input matrix of dimension [200*8000] I have output of [1*1]. I used Neural fitting toolbox (nftool) of MATLAB but it was giving me error by saying " Data selections have different number of samples ". So I want to know how to build mapping between input and output so later on I will give new input to that mapping to get new output?
Hi,
if you have an input matrix of predictors of size [200x800000] your output vector should be [200x1], independently on the dimensionality of the input space. Are you sure that you do not have 800000 samples per 200 input dimensions? Having 200 samples per 800000 dimensions as you described seems a very high dimensionality space with very few samples.
Samples are Matrix columns, so I have 800000 samples of 200 . but my problem is my output is 100 samples of 1. So how to perform mapping?
- Badges
- Science topic