Given two different graphs, can I use SVM to detect whether they are closely (not exactly) matching with a defined margin of error or threshold? If yes, what are the steps? How do I do it? Sorry I am very new to the field of machine learning and appreciate expertise help.
Reason I am asking is that I have a set of (x) inputs over (t) time that I would like to verify and match against predefined sets of (x) values over (t). This can be used in motion type detection using accelerometers on mobile devices.
Taking the the first sentences of your two paragraphs: you just want to detect when a new
instance is “close to” any one of the instances in your predefined set of time-series.
The straightforward way to do that is to just do it; no need for fancy machine learning algorithms.
If you don’t want to compare each instance against everything in you predefined set,
then you could (possibly) try to distil it down to a fewer number prototypes, using a clustering or other unsupervised learning algorithm.
The SVM is typically used to solve data-driven classification problems where you have:
two (or more) labelled data sets, each with many instances; each instance has a set of feature values, and you want to construct a classification model that will label subsequent data.