Neurocomputing article by LIACS researchers most cited
LIACS scientist, Dr. Michael S. Lew, along with his students now has the most cited article in the top tier neural network journal Neurocomputing, the second highest Google Scholar impact over all neural network journals worldwide.
New paradigms and ideas
In "Deep Learning for Visual Understanding" the authors describe important new paradigms and ideas such as recent neural network architectures and novel approaches for training neural networks. They also describe auto-encoders which can learn to computationally synthesize new examples, much like the human imagination.
Then they discuss and propose the important research directions for the future, such as using feedback from other networks and how to integrate multiple modalities of information. Moreover, they explain the role of limited training set data and the lack of insight into the deep models.
For more information, visit this link.