Summary:
In the current study, we consider the problem of adaptive learning in the field
of Machine Learning together the respective algorithms. In adaptive learning,
the goal is to study and develop algorithms, which are capable of learning
through input/output measurements, and adapt in changes of the environment, in
order to estimate unknown parameters at each time instant. Such problems are
encountered in a lot of practical applications, such as mobile communications,
and it characterizes systems that work in real time. Thus, in the respective
literature, several adaptive algorithms have been studied, that achieve a good
estimation of unknown parameters. At the study of adaptive algorithms, it
becomes easily perceptible the need for the existence of important and
extensive theoretical backgrounds for their creation and their operation. There
are many algorithms that are used for linear but also for non linear systems,
applying intelligent mathematic techniques. Concerning the non linear systems,
the development of adaptive algorithms is more complicated, but, with a
suitable adaptation in the mathematic models, this may lead to very good
results. In the current manuscript, we will study the philosophy of adaptive
algorithms, their theoretical background, where we will discuss them in detail,
referring concepts such as cost function, kernels, and, of course, in the end
we will see their experimental implementation, comparing the various
experiments of algorithms and exporting conclusions.
Keywords:
adaptive learning, adaptive algorithms, linear and non linear systems, convex sets, kernels
