I studied physics at the Technical University of Darmstadt from 2011 to 2016. For the work on my master’s thesis, I joined the lab of Jochen Triesch at the FIAS Frankfurt. In my thesis, I used a self-organizing recurrent neural network to model sequence learning in primary visual cortex. Since 2017 I am part of the Neural Network Dynamics and Computation group of Raoul-Martin Memmesheimer at the University of Bonn, first as a PhD student and since 2022 as a Postdoctoral fellow.
Broadly speaking, I am interested in the computational capabilities and dynamical features of neural network models. More specifically, I am interested in learning algorithms that lie at the intersection of neuroscience and machine learning, such as learning without weight changes (dynamical learning) and perturbation-based learning. To understand their functioning, I use methods from physics. In addition, I’ve also worked on the problem of how memories can persist despite ongoing connection weight changes and on the effects of epilepsy-induced changes on network function. The latter project was done in collaboration with the experimental lab of Heinz Beck.
P. Züge, C. Klos, and R.-M. Memmesheimer (2021)
Weight perturbation learning outperforms node perturbation on broad classes
of temporally extended tasks
Y.F. Kalle Kossio, S. Goedeke*, C. Klos*, and R.-M. Memmesheimer (2021)
Drifting assemblies for persistent memory: Neuron transitions and unsupervised compensation
Proc. Natl. Acad. Sci. USA, .
C. Klos, Y.F. Kalle Kossio, S. Goedeke, A. Gilra, R.-M. Memmesheimer (2020)
Dynamical learning of dynamics
Phys. Rev. Lett. 125:088103
, , , , , (2019)
Altered dynamics of canonical feed-back inhibition predicts increased burst transmission in chronic epilepsy
J. Neurosci. 2594-18.
C. Klos, D. Miner and J. Triesch (2018)
Bridging structure and function: A model of sequence learning and prediction in primary visual cortex
PLoS Comput. Biol. 14 (6): e1006187.