If we have a SISO dynamical system, for reduced order approximation using ML algs, do we need to take (frequency/ time domain) samples of Impulse response/TF
Then to approximate this input/output data using reduced model?
used correlation, the highest one you can remove from your data...for reducing variables
Dear Shakir Sofi,
Dynamic systems produce time series, so you can implement some types of data fusion for enabling data reduction, the following paper do the same for IoT time series coming from similar smart things (same cluster)
https://www.researchgate.net/publication/346060936_Data_Reduction_Model_for_Balancing_Indexing_and_Securing_Resources_in_the_Internet_of_Things_Applications
To reduce data you have a number of different options:
1- Feature extraction techniques (Using different techniques with different strategies (PCA, mean, max, min, kurtosis, skewness and etc.))
2- Feature selection techniques (DT, ICA, LDA, evolutionary algorithms such as PSO, GA, ACO,...)
- Badges
- Science topic