- Master Thesis
Rights / licenseIn Copyright - Non-Commercial Use Permitted
Backpropagation is the engine driving the success behind modern artificial neural networks. In light of the recent slow-down of breakthroughs, researchers are increasingly looking towards the brain for renewed inspiration. We extend the Deep Feedback Control framework (1) to implement a novel network with recurrent connections using a local learning rule that still fits into the input-output paradigm of modern machine learning. The further extension to time-series data did not yield any significant success, which is thoroughly analyzed on a theoretical and practical level. This thesis also provides the translation of the neuro-scientific idea of neural assemblies into a computational paradigm. Show more
Organisational unit09479 - Grewe, Benjamin F. / Grewe, Benjamin F.
MoreShow all metadata