Efficient Deep Learning programme kick-off

© ASTRON

Artificial intelligence and in particular deep learning are increasingly becoming part of our society, either embedded in our infrastructure and day to day products, or present in our interactions with machines. Examples are voice-text conversion, translation, object recognition, etcetera.

Deep learning is an advanced form of artificial intelligence, with neural nets containing many layers, able to learn without requiring manual preprocessing steps. However, a big bottleneck is training these neural nets. The nets usually are complex, requiring lots and lots of compute power, and also requiring a very large number of training data sets. These data sets need to be labeled, and creating these labels is often very time consuming and expensive.

In order to tackle these and other deep learning challenges, the NWO Perspectief programme Efficient Deep Learning (EDL) started a broad coherent research programme consisting of seven projects ad six research lines, binding over 40 universities, knowledge institutes, and companies. On January 17/18 the kick-off was held in Soesterberg.

The photo shows the EDL programme leader prof. Henk Corporaal (TuE) explaining deep learning and the main programme goals. ASTRON participates in the project Deep Learning for HTSM, led by prof. Rob van Nieuwpoort (UvA, Netherlands Escience Center), and contributes the use case LOFAR system health.