These two I find really fascinating: the brain and human society. From molecules over neurons to brain areas, from individuals over organisations to nations, both are made up of countless moving parts, span orders of magnitude and constantly evolve. With a Physicist's mindset, I've squinted my eyes, blending out many details but trying to find simple principles that describe these otherwise hopelessly complex systems. The nature of stability and decentralised information processing are such principles that I've worked on.
Information processing in the brain
Currently, I'm a 3rd year PhD student at the Institute of Neuroinformatics in Zurich in Jean-Pascal Pfisters lab. The bigger picture of our research is understanding how the brain processes information. For survival, knowing what you don't know is arguably more important than what you know. So my part in this is the brain's handling of uncertainty. I've designed an experiment to test overall brain function but I've also borrowed ideas from machine learning to explain how synapses, the connections between neurons, handle uncertainty. One of the most amazing things in neuroscience is how well tuned everything works together without any of the bits and pieces understanding how anything they do fits into the grand scheme of things.
Does stochastic synaptic transmission implement Bayesian regression? To answer this question, we conduct a big thought experiment: we start with the assumption that our hypothesis is true and derive closed form plasticity rules for spiking perceptron like networks from there. We can then validate these plasticity rules with electrophysiological recordings, like the STDP data to the left.
Is the brain capable of Bayesian Regression? Together with Michael Herzog's lab at EPFL, we designed a psychophysics experiment and our results are positive. Since Bayesian Regression involves some complicated math we think that our experiments adds another constraint on the neuronal coding of uncertainty. Draft coming soon!
Stability and environmental science
During my Master's Adrian and I extended a method to compute stability for broad range of models that can be used in ecosystems but also for memories in brain or electrical power grids - as long as the system can be cast into a sufficiently abstract mathematical model. Previously researchers had only asked if a system could ever recover from a large schock - we asked (and answered) how to account for all the different pains the system might suffer while recovering. I've also contributed to the methods section of work done by Julian, trying to understand the role of photophore storage in soil.
Adrian van Kan*, Jannes Jegminat*, Jonathan F. Donges, and Jürgen Kurths
*(joint first authorship)
Phys. Rev. E – 2016
Julian Helfenstein, Jannes Jegminat, Timothy I. McLaren and Emmanuel Frossard
Biogeosciences – 2018