MIMO system identification and Correlation among inputs

I want to estimate Non-Parametric and Parametric models based on MIMO signals (2 inputs and 5 outputs).

When i estimate the correlation between the two inputs, the result is high correlation.

What can i do to tackle this problem? What are the possible tools?

Answer 1

[Balavignesh]

Hi Dionysis,

As per my understanding, you would like to estimate non-parametric and parametric models based on MIMO signals, but there is high correlation between the two inputs.

To tackle the problem of high correlation between the two inputs in MIMO system modeling using MATLAB, I would suggest you follow any of the following techniques.

• Regularization: MATLAB provides functions like 'ridge' and 'lasso' for regualrized regression. These functions can be used to introduce a penalty term and control the impact of high correlation during model estimation.
• Nonlinear Modeling: MATALB provides tools for non-paramteric modeling, such as neural networks using the 'nftool' or 'fitnet'. These tools can capture complex nonlinear relationships in the MIMO systems.

Please refer to the following documentation links to have more information on:

• 'ridge' function: https://in.mathworks.com/help/stats/ridge.html
• 'lasso' function: https://in.mathworks.com/help/stats/lasso-regularization.html
• 'fitnet' function: https://in.mathworks.com/help/deeplearning/ref/fitnet.html

Hope that helps!

Balavignesh