For example, building data models (e.g., clustering or frequent patterns) in high speed data streams require the use of machine learning techniques in AI to fix the problem of concept drifts and time-variations. In many scientific applications, where data is distributed and large, the concept of utility in AI is used to evaluate the cost of data mining tasks (e.g., data acquisition, data mining, and model utilization) so that knowledge discovery is practically feasible in resource constrained environments. Agent-based techniques are now used to reason and coordinate knowledge discovery tasks across distributed data repositories; neutral networks are now used to optimize data mining parameters; spectral clustering is now used in case-based reasoning (CBR) for medical discovery, and incremental learning is now the means to "idiot-proof" business intelligence systems.

Both disciplines have carried this potential to be used in a closed-loop fashion, where the reasoning of AI helps to "soften" the problems brought about by the brute force analytics of data mining. And data mining in turn, is the key to producing the relevant models and patterns that AI algorithms require. As users expect more from intelligent systems, there is further motivation for researchers of both disciplines to exploit the possibilities of what this closed-loop framework can potentially offer.

The objective of this workshop is to collect and report the experiences by researchers in either or both disciplines, and to offer an opportunity for researchers around the world to meet and share their ideas. The second run of this workshop received a total of