Feedforward Net for Classification Rank features

Hi dear community,

I'm working with a Feedforward Backpropagation Network to classify 5 labels. The Net has 30 input parameters which represent such a deformed signal and 5 labels to classify the type of deformation. The number of neurons in the hidden layer was chosen for the best performance.

Generate pattern recognition network - MATLAB patternnet (mathworks.com)

Trainning Database:

[30x5000] Matrix, 5000 columns with 30 features each, where 1000 data per target (signal distortion type)

I'm needing to rank the 30 features, I checked built in function in Matlab to do that, and I used

-----------------------

Introduction to Feature Selection - MATLAB & Simulink (mathworks.com)

relieffClassification and regressionEither all categorical or all continuous features

Rank features using the ReliefF algorithm for classification and the RReliefF algorithm for regression. This algorithm works best for estimating feature importance for distance-based supervised models that use pairwise distances between observations to predict the response.

-----------------------

% Target Label Loading
load('C:\Users.......\Target_LabelNumbered.mat');

% Trnsponse datasets before valuating by relieff
[idx,weights] = relieff(TestMat_Sinc_25dB',Target_LabelNumbered',5);
bar(weights(idx))
xlabel('Predictor rank')
ylabel('Predictor importance weight')

The function works fine with the database and labels, and it easy to use.

But I'm wondering to know if using this algorithm with k-nearest method suits well to rank the features that are used on a Feedforward Backpropagation Classifier? or should I try a different kind of test using the trainned network instead?

Thanks in advanced!