Abstract: Learning from data streams originating from non-stationary environments is vital for many real-world applications. A notable challenge in this task is concept drift. Most existing methods ...