NWO subsidy for efficient deep learning systems
Computer scientists from the UvA, VU and NLeSC received a subsidy from NWO Perspectief for the Efficient Deep Learning program.
Researchers Henri Bal, Mark Hoogendoorn (VU), Theo Gevers, Cees Snoek and Max Welling (UvA) and Rob van Nieuwpoort (NLeSC/UvA), from Artificial Intelligence and Computersystems will study efficient self-learning systems for applications like video surveillance, 3D reconstruction and steel inspection, together with companies like Schiphol, 3DUniversum, ING and Tata Steel. A service is also being set up that makes it easier for companies to use self learning systems. With the grant, seven PhD students and a postdoc are appointed. The duration of the project is five years.
Efficient deep learning
A computer that recognises dangerous situations on security footage: this is possible with deep learning automated systems. But before this kind of system can operate independently, you have to design it and then train it with a huge number of examples. In addition, you need considerable computing power to let the system make decisions. At the Efficient Deep Learning programme, researchers are going to make deep learning much more efficient by using examples from daily life. They want to make it possible to use the technique (Of: They want to make the technique applicable) for other automatic visual inspections, tissue analysis, smart maintenance of equipment and intelligent hearing aids that can handle noisy environments.
With Perspectief, NWO is challenging scientists to establish a close partnership with industry and social organisations. It concerns multidisciplinary research with a special emphasis on application. Together the parties will develop new research lines linked to the top sectors. A research programme generally consists of five to six projects that approach the overarching theme from different research disciplines. The board of NWO Domain Applied and Engineering Sciences (AES) is providing 21 million euros for Perspectief.
Program leader is Prof.dr. H. Corporaal (Technische Universiteit Eindhoven) and the partners are: AIIR Innovations, ASTRON, CWI, Cyclomedia, Cygnify, Donders Instituut, FEI, 2getthere, GN Hearing, Holst Centre, ING, Intel, Irdeto, Lely, Mobiquity, NLeSC, NXP, NVIDIA, Océ, Radboudumc, Schiphol, Scyfer, Sectra, Semiotic Labs, Siemens, Sightcorp, Sorama, SURFsara, TASS International, Tata Steel, Technische Universität Dresden, Technische Universiteit Delft, Technische Universiteit Eindhoven, Thales, TNO, TomTom, Universiteit Twente, Universiteit van Amsterdam, 3DUniversum, VicarVision, ViNotion, Vrije Universiteit Amsterdam, Wageningen University & Research.