Machine, learn quicker!

Symbolic picture for the article. The link opens the image in a large view.
Bild: Colourbox

Scientists at FAU develop a method for accelerating deep learning processes.

Deep learning involves machines learning using neural networks. However, they have to learn from scratch each time they are used in a different area. Researchers at Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) have now found a way of accelerating these learning processes by implementing prior knowledge in the networks without the need for time-consuming learning processes. They have published their findings in the journal Nature Machine Intelligence.

This is a break from the traditional procedure which dictates that all information used to train the networks have to come from the data provided for the task in question, and is a revolutionary concept in computer science. This approach has even allowed the scientists led by Prof. Andreas Maier, Chair of Computer Science 5 (Pattern Recognition) to better understand the complex deep neural networks. Although computer scientists have been developing neural networks successfully for years now, it is still not entirely clear how these networks actually work.

Further information

Prof. Andreas Maier is also involved in the European Time Machine project:

Prof. Andreas Maier is one of the FAU scientists researching artificial intelligence. Please refer to the FAU AI website for an overview of the researchers and their AI projects.

Prof. Dr. Andreas Maier
Phone: +49 9131 85 27883
andreas.maier@fau.de