Can we say that system identification is almost same thing as Machine learning or they both different techniques? I am confused because as shown in attached snap of matlab website, system identification Create linear and nonlinear dynamic system models from input-output data which is same as what happens in Machine learning, Model is obtained from input-output data
Can we say that system identification is almost same thing as Machine learning or they both different techniques?
They clearly have different scopes: you can identify a system with non-learning methods, and machine learning can do other things than identifying systems.
"A duck eats bread" and "Marcus is a bread-eating organism" doesn't say "a marcus and a duck are the same".
Machine learning starts with an assumption that you have a lot of assorted data, you don't know what you're doing, and you're going to let the computer do all the work*.
System identification starts with an assumption that you have a lot of very specific data, preferably that has been captured with system ID in mind, and that you have a pretty good idea of the structure** of your system to start with.
A lot of the underlying math is common between the two, but machine learning tends to compromise on overcoming a lack of knowledge of the system to be modeled with tons and tons of data, while formal system ID tends to compromise on overcoming a lack of tons and tons of data with some prior knowledge of the system to be identified.
* For some machine learning efforts this is an exaggeration -- but, unfortunately, not all.
** "Structure" in this case meaning how many states, where the nonlinearities are, what sort of nonlinearities there may be if not their exact nature, etc.
