Balanced excitation and inhibition are required for high-capacity, noise-robust neuronal selectivity

Abstract

Neurons and networks in the cerebral cortex must operate reliably despite multiple sources of noise. To evaluate the impact of both input and output noise, we determine the robustness of single-neuron stimulus selective responses, as well as the robustness of attractor states of networks of neurons performing memory tasks. We find that robustness to output noise requires synaptic connections to be in a balanced regime in which excitation and inhibition are strong and largely cancel each other. We evaluate the conditions required for this regime to exist and determine the properties of networks operating within it. A plausible synaptic plasticity rule for learning that balances weight configurations is presented. Our theory predicts an optimal ratio of the number of excitatory and inhibitory synapses for maximizing the encoding capacity of balanced networks for given statistics of afferent activations. Previous work has shown that balanced networks amplify spatiotemporal variability and account for observed asynchronous irregular states. Here we present a distinct type of balanced network that amplifies small changes in the impinging signals and emerges automatically from learning to perform neuronal and network functions robustly.

Robust associative learning is sufficient to explain structural and dynamical properties of local cortical circuits

“…According to this measure, neurons exhibit greater irregular activity when memory load and rescaled robustness are high, with CV of ISI values saturating at around 0.9. This is consistent with the range of CV of ISI values reported for different cortical systems, 0.7-1.1 [5][6][7][8][9] . To examine the extent of synchrony in neuron activity we calculated spike train cross-correlation coefficients for pairs of neurons ( Figure 4C).…”

“…In the biologically plausible region of parameters, individual neurons are loaded with relatively long sequences of network states (0.2N for the green asterisk in Figure 5A), yet it is not clear if the associations learned by individual neurons assemble into memory sequences that can be successfully retrieved at the network level. 9 To examine this question we first tested memory retrieval in the absence of noise. For this, we initialized the network state at the beginning of the loaded sequence and monitored playout of the memory.…”

“…These predictions are consistent with a large number of experimental studies of connectivity in local cortical circuits. Other models, including only some of the constraints considered in this study, have been previously described [4][5][6][7][8][9] .…”

“…For example, individual neurons exhibit highly irregular spiking activity, resembling Poisson processes with close to one CV in inter-spike-intervals [5][6][7][8][9] . Spike trains of nearby neurons are only 4 connections of each neuron are homeostatically constrained to have a fixed average absolute weight, w [28][29][30] .…”

The effects of individualised intermittent theta burst stimulation in the prefrontal cortex: A TMS‐EEG study

“…This ratio may represent excitation/inhibition balance in the prefrontal cortex(Noda, Zomorrodi, Cash, et al, 2017d) which is important in working memory and network functions(Legon et al, 2016;Rubin, Abbott, & Sompolinsky, 2017), and balanced regulation is required for prefrontal cortex-dependent behaviours(Fan & Hu, 2018). This ratio may represent excitation/inhibition balance in the prefrontal cortex(Noda, Zomorrodi, Cash, et al, 2017d) which is important in working memory and network functions(Legon et al, 2016;Rubin, Abbott, & Sompolinsky, 2017), and balanced regulation is required for prefrontal cortex-dependent behaviours(Fan & Hu, 2018).…”