Geometry and dynamics of representations in a precisely balanced memory network related to olfactory cortex

Published: Dec. 31, 2024

Biological memory networks are thought to store information by experience-dependent changes in the synaptic connectivity between assemblies of neurons. Recent models suggest that these contain both excitatory and inhibitory neurons (E/I assemblies), resulting co-tuning precise balance excitation inhibition. To understand computational consequences E/I under biologically realistic constraints we built a spiking network model based on experimental data from telencephalic area Dp adult zebrafish, precisely balanced recurrent homologous piriform cortex. We found stabilized firing rate distributions compared with global Unlike classical models, did not show discrete attractor dynamics. Rather, responses learned inputs were locally constrained onto manifolds “focused” activity into neuronal subspaces. The covariance structure supported pattern classification when was retrieved selected subsets. Networks therefore transformed geometry coding space, continuous representations reflected relatedness an individual’s experience. Such enable fast classification, can support continual learning, may provide basis for higher-order learning cognitive computations.

Language: Английский

The topology of E/I recurrent networks regulates the effects of synaptic plasticity

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Feb. 28, 2023

Cortical neurons are versatile and efficient coding units that develop strong preferences for specific stimulus characteristics. The sharpness of tuning efficiency is hypothesized to be controlled by delicately balanced excitation inhibition. These observations suggest a need detailed co-tuning excitatory inhibitory populations. Theoretical studies have demonstrated combination plasticity rules can lead the emergence excitation/inhibition (E/I) cotuning in driven independent, low-noise signals. However, cortical signals typically noisy originate from highly recurrent networks, generating correlations inputs. This raises questions about ability mechanisms self-organize co-tuned connectivity receiving noisy, correlated Here, we study input selectivity weight neuron network via plastic feedforward connections. We demonstrate while noise levels destroy readout neuron, introducing structures non-plastic pre-synaptic re-establish it favourable correlation structure population activity. further show structured impact statistics fully driving formation do not receive direct other areas. Our findings indicate dynamics created simple, biologically plausible structural patterns enhance synaptic learn input-output relationships higher brain

Language: Английский