prof. dr. Lambert Schomaker
Artificial Intelligence and Cognitive Engineering (ALICE)
[Research | Education]
Bernoulliborg building (V)
9747 AG Groningen, The Netherlands
E-mail: home email address
Interests & projects
- Within artificial intelligence, my focus is on perceptual intelligence and (lifelong) machine learning
- Monk is our continuous project for allowing access to large and diverse
historical archives which are difficult to access by means
of traditional 'OCR' approaches due to their special fonts or handwritten style.
Since 2009, Monk provides a means of continuously training handwriting recognizers that create indices of handwritten
collections. Monk was one of the demonstrators in the EU/SNN Target project
- May 2019: EU/ITN Manic project approved, participation: ML and neuromorphic computing / cognitive materials
- 2018 EU/ITN Perico project participation: Machine learning for peroxisome classification
- Dead Sea Scrolls ERC project with Mladen Popovic (PI) and Hans van der Plicht (14C dating)
- 2021-06-01 - Currently developing: the HAICu project on AI for the cultural heritage
- 2020-09-08 - PhD position for AI student with physics interest
(due to EU regulations this position is not intended for UG students, sorry. Candidates need to have received
their training in AI elsewhere). The scientific aim of MANIC is to synthesize materials that can function as
networks of neurons and synapses by integrating conductivity, plasticity and self-organization.
- 2020-03-03 - AI colloqium From Boston to Eden - or how to get systems that are really autonomous and sufficiently intelligent to survive in their niche.
In this presentation I argue that reinforcement learning methods currently are still very much handcrafted. Only an optimal
reward structure and schedule yields success, which requires often more than an single PhD project. On the other hand,
the human and animal brain do not have the helping hand of an ever-present external engineer setting up the framework for survival.
Modern AI should take a better look at what is known about the old brain, which provides a general but quick valuation of
the current state of the world. This would allow for a really autonomous reinforcement learning which does not require the development
of painstakingly handcrafted reward structures, neural-network hyper-parameter values and specifically optimized training schedules.
- Older Publications,
- People in my group
- Former M.Sc. Students
- Ph.D. involvement
Current PhD students
- Zhenxing Zhang - Visual Question Answering architectures
- Asmaa Haja - Deep learning methods for microscopy in biology
- Anouk Goossens - Building blocks for electronic brains (co-supervisor with prof. Tamalika Banerjee)
- Sha Luo - Reinforcement-based learning of object handling in robots
- Mahya Ameryan - Deep learning for text recognition and pictorial semantics
- Maruf Dhali - Dead Sea Scrolls, document dating and writer identification
- Jean-Paul van Oosten - Continuous labeling and learning
Jean-Paul has obtained, at the ICFHR 2012, the IAPR Best Paper Award:
Jean-Paul Van Oosten, Lambert Schomaker (2010).
Separability versus Prototypicality in Handwritten Word Retrieval,
Proc. Int. Conference on Frontiers in Handwriting Recognition, September 18-20 2012, Bari, Italy, IEEE Computer Society, pp. 8-13,
- Amir Shantia - Robotics perception and navigation (together with Marco Wiering)
(previous PhD students)
- Bowornrat Sriman (camera-based text detection and recognition for Asian scripts)
- Sheng He (writing style-based dating of handwritten manuscripts, dissertation 2017, 'cum laude' - with honors)
- Marius Bulacu (writer identification, dissertation 2007)
- Tijn van der Zant (handwritten historical document retrieval, dissertation 2010)
- Axel Brink (robust writer identification and verification, dissertation 2011)
- Gert Kootstra (biologically inspired robot vision and navigation, dissertation 2010)
- Hado van Hasselt (reinforcement learning systems, with Marco Wiering, dissertation 2011)
Setare Rezaee is a PhD guest, working on Persian (Farsi) handwriting recognition using deep learning (2018/2019).