Excitatory/inhibitory balance emerges as a key factor for RBN performance, overriding attractor dynamics

Abstract

Reservoir computing provides a time and cost-efficient alternative to traditional learning methods. Critical regimes, known as the “edge of chaos,” have been found to optimize computational performance in binary neural networks. However, little attention has been devoted to studying reservoir-to-reservoir variability when investigating the link between connectivity, dynamics, and performance. As physical reservoir computers become more prevalent, developing a systematic approach to network design is crucial. In this article, we examine Random Boolean Networks (RBNs) and demonstrate that specific distribution parameters can lead to diverse dynamics near critical points. We identify distinct dynamical attractors and quantify their statistics, revealing that most reservoirs possess a dominant attractor. We then evaluate performance in two challenging tasks, memorization and prediction, and find that a positive excitatory balance produces a critical point with higher memory performance. In comparison, a negative inhibitory balance delivers another critical point with better prediction performance. Interestingly, we show that the intrinsic attractor dynamics have little influence on performance in either case.

The connectivity degree controls the difficulty in reservoir design of random boolean networks

“…In particular, Metzner and Krauss ( 2022 ) suggested a more complex picture than previously thought, exposing two critical points, each for a positive and negative balance, while for higher densities, an asymmetry could arise in the reservoir responses to inputs, and as a result, only the edge of chaos occurring for positive b was optimal for information propagation inside the reservoir. In line with Krauss and Metzner, recent work on RBN reservoirs demonstrated that the excitatory-inhibitory balance b was also key in driving dynamics and performance (Calvet et al, 2023 ). In particular, it was shown that the weight statistics, typically used in RBN literature (Bertschinger and Natschläger, 2004 ; Natschläger et al, 2005 ; Büsing et al, 2010 ) are related to the balance.…”

“…In this article, we want to study the effect of these topology parameters ( N and K ) with another control parameter, less studied in this context, which is the excitatory-inhibitory balance b , controlling the proportion of positive and negative synaptic weights (Krauss et al, 2019a ; Metzner and Krauss, 2022 ; Calvet et al, 2023 ). More specifically, the balance is equal to b = ( S + − S − )/ S , with S = KN the total number of synapses and S ± the number of positive and negative synapses.…”

“…Primarily, this balance is fundamental to the principle of homeostasis, which prevents the brain from overflowing with spikes and keeps the average activity in a certain range (Sprekeler, 2017 ). It has been shown that strong excitation can provoke irregular activity patterns (Van Vreeswijk and Sompolinsky, 1996 , 2005 ; Krauss et al, 2019b ; Sanzeni et al, 2022 ; Calvet et al, 2023 ), and that an imbalance of excitation and inhibition could be linked to pathologies such as epilepsy (Nelson and Valakh, 2015 ) and autism (Arviv et al, 2016 ). In our present context, studies on models (Ehsani and Jost, 2022 ), in vitro (Sandvig and Fiskum, 2020 ) and in vivo (Yang et al, 2012 ), showed that the meticulous balancing of excitatory and inhibitory neurons was also linked to the edge of chaos (Poil et al, 2012 ).…”

“…Despite its importance in neurosciences, the excitatory-inhibitory balance has only been recently introduced for investigating the design of RBN (Calvet et al, 2023 ). Previous work on the ESN (Krauss et al, 2019a , b ; Metzner and Krauss, 2022 ) has studied the influence of density d = K / N and balance on the dynamics of reservoirs, showing that b was a key parameter controlling phase transitions.…”

“…As far as the authors are aware, the influence of the excitatory-inhibitory balance for different connectivity has yet to be studied, except for the single value of K = 16 previously mentioned (Calvet et al, 2023 ). This article aims to explore the combined effect of connectivity ( K , N ) and the balance on the dynamics and performance of the RBN.…”