Deep learning uses neural network architectures with many layers of neurons. This allows the network to learn abstract concepts, and such networks are used successfully in areas like image classification and object recognition.
The deep learning revolution started in 2012, when Krizhevsky et al. used a deep convolutional network to classify images from the ImageNet database with unprecedented performance. Since then, deep learning has been successfully applied in many different fields, and it is one of the most central technologies in addressing current big data challenges.
We collaborate closely with the Institute of Marine Research to apply deep learning techniques large data sets collected from the marine domain. This includes image data, but also data from oceanography, acoustics (sonar and echo sounder), and lab generated data. The lack of good training sets is a challenge, and unsupervised and semi-supervised methods are therefore especially important.