Linear Time Invariant State Space System Identification Using Adam Optimization
In this paper, linear time invariant systems are modelled in state-space as Tensorflow graphs.
Given the systems' inputs and outputs, Adam optimiser is used to optimise state space matrices.
Some assumptions were made to build the model. The technique was then tested and was proven to
work even with high noise to signal ratios.
Moreover, the technique successfully identified a system that didn't follow all the rules from
the assumptions the model was built on, showing great potential, despite the simplicity of this
technique.
The paper is on IEEEXplore and I have published the code for this paper and some
examples on my GitHub.