PhD position (4 years, fully funded) in Statistical Physics of Learning

Updated: over 1 year ago
Deadline: 2022-09-01T00:00:00Z

Two fully funded (4 years) PhD positions in  Machine Learning / Deep Learning are available at the University  of Groningen within the project:  "Robust Learning of Sparse Representations: Brain-inspired Inhibition and Statistical Physics Analysis".


Main supervisors and sub-projects:  

WP1) George Azzopardi (https://www.cs.rug.nl/~george/)

"Push-pull inhibition in deep, convolutional (spiking) neural networks"


WP2) Michael Biehl (https://www.cs.rug.nl/~biehl/) 

"Statistical physics analysis of learning processes in model situations" 


WP2 will focus on the systematic study of learning processes in model situations. To this end, analytical methods borrowed 

from statistical physics will be applied to large layered systems of model neurons. We will focus on the consideration of activation 

functions which (a) are inspired by or relate to realizable systems for neuromorphic computing and (b) favor sparse activity in the 

working phase. In addition, training algorithms and loss functions will be investigated that favor or enforce sparse connectivity. 


The WPs are complementary to each other in that they contribute to the repeating pair of the core building blocks in deep architectures of deep architectures, namely convolutions (WP1) and activation functions (WP2).  Understanding the interplay and integration of these two concepts in one system will be targeted in joint efforts of the two WPs throughout the project. 


For WP2 we are looking for a candidate with a strong background in statistical physics and a deep interest in the application, 

modelling and theory of neural networks and machine learning. 


For further details and expression of interest please visit  https://t.co/VLjgKzPxJp

where you should submit your expression of interest as soon as possible. An official application procedure

will be established soon. 



Similar Positions