Neuronal activity patterns change continuously over days and weeks, a phenomenon known as representational drift. Despite this, the geometric structure of population representations, namely the pairwise similarities between stimulus-evoked activity patterns, remains remarkably stable. How can ongoing changes in activity be consistent with stable representational similarit? We show that this is a generic consequence of random connectivity: in networks with random connectivity, output similarity is a monotonically increasing function of input similarity, independent of the specific connectivity pattern. Drift, whether driven by random synaptic turnover or Hebbian plasticity, merely transitions the network between random instantiations, leaving similarity intact. This extends to recurrent architectures and to deep neural networks, where continued training beyond performance saturation produces activity drift while preserving representationalsimilarity.
Although connectivity in the brain is not random, networks trained on high-dimensional inputs acquire connectivity that behaves statistically like a random projection, making these results broadly applicable to biological neural circuits.

About the Speaker: Alex Roxin is Group Leader in the Computational and Mathematical Neuroscience unit at the Centre de Recerca Matemàtica (CRM, Barcelona). “Research in the computational neuroscience unit is largely focused on systems-level neuroscience. This involves investigating how large assemblies of interacting neurons give rise to animal and human behavior. Our approach is generally to combine computational modeling with data analysis.” (Source, CRM)
Read more about Alex Roxin and work in his team – https://www.crm.cat/_neuroscience/

Photo courtesy of Alex Roxin

This talk is hosted by Denis Alevi (Associated PhD B03)

Certificate of attendance: Please contact team assistant Laura.Starck@hu-berlin.de

Series Overview