Machine intelligence concerns the analysis, design, and implementation of intelligent agents/systems. These types of systems appear in a wide and diverse range of applications, and includes recommender systems, systems for troubleshooting faulty equipment (and humans) as well as systems for doing process and customer monitoring and profiling. Common for these applications is that the systems must reason and operate under uncertainty (due to, e.g., incomplete domain knowledge or noisy sensors). Furthermore, the systems should be able to learn from experience (machine learning). Here experience may take the form of past information collected in a database or observations continuously being generated by the environment in which the system operates.
The machine intelligence activities in CISS are mostly centered around probabilistic and statistical models, which provide a principled foundation for frameworks for learning and reasoning under uncertainty. The algorithms that have been developed by the unit target both classical machine intelligence settings, where data consists of independent observations (as typically found in flat data files), as well as data with a more rich structure capturing possible relations among the observations (as one finds in relational databases). The unit has a long and established position in this research field, where the developed algorithms have been applied for, e.g., monitoring of production plants, predicting customer solvency, and doing traffic maneuver recognition.