Marcel Beekveld [Thales]

Machine learning in the naval domain

10:15 – 10:55 and 14:00 – 14:40 (subject to change)

At Thales Netherlands, machine learning and AI are being applied in more and more areas and have also found their way into commercial products. The specific challenge for some applications is that parameters in models depend on the specific nature of naval missions and the geographical location. Although machine learning seams ideally suited to train systems for varying situations, data collection in the naval domain comes with its own specific problems. Mission data is not only dependent on the geographical location, data is also sparse, temporally inconsistent, and the data sets are relatively small. Also, data may be confidential to some extent. To overcome this, special measures are being taken, such as data augmentation, smart resolution of temporal inconsistencies and data anonymization without loss of information.

Marcel Beekveld has been active in a number of areas as software architect including the machine industry, mechatronics and defence. Since 2008 he has been working at Thales Netherlands in a number of smaller and larger defence-related projects, as well as innovation projects. The latest innovation projects have involved automated classification of objects based on naval sensor data.