Multimedia data is produced in numerous formats. Hence, the
extraction of information from these data is becoming ever more challenging. Magnifying this is that information must frequently be extracted from many distinct and often distributed data streams and it must frequently be
extracted in real time to be of value.
I start by outlining these problems, which have led
to the emergence of a new and exciting field of research: stream mining. Next,
I argue that this field requires the development of novel processing, learning,
networking, and computing paradigms. I conclude by presenting our efforts in
building this field in the past 7 years.