Topic: recurrent spiking networks

Information encoding in neural circuits depends on how well time-varying stimuli are encoded by neural populations. Slow neuronal timescales, noise, and network chaos can compromise reliable and rapid population response to external stimuli. A dynamic balance of externally incoming currents by strong recurrent inhibition was previously proposed as a mechanism to accurately and robustly encode a time-varying stimulus in balanced networks of binary neurons, but a theory for recurrent rate networks was missing. Here, we develop a non-stationary dynamic mean-field theory that transparently expla

Synaptic plasticity is known to be a key player in the brain’s life-long learning abilities. However, due to experimental limitations, the nature of the local changes at individual synapses and their link with emerging network-level computations remain unclear. I will present a numerical, meta-learning approach to deduce plasticity rules from either neuronal activity data and/or prior knowledge about the network's computation. I will first show how to recover known rules, given a human-designed loss function in rate networks, or directly from data, using an adversarial approach. Then I will pr

Synaptic plasticity is thought to underlie learning and memory [1]. Plasticity rules can be active in different synaptic connection types, such as excitatory-excitatory (EE) and inhibitory-excitatory (IE) connections, with in silico and experimental evidence suggesting distinct mechanisms for each type [2,3,4]. While the interaction between connection types in learning and memory is an active research focus [5], studies typically probe only one type of plasticity at a time. Thus, at scale, the contributions of connection types and their co-active plasticity rules to memory processes remain unc

Recent theoretical models and experimental data have revealed that many neurons can exhibit homoclinic (HOM) spike-onset bifurcations by tuning variables like temperature, extracellular ion concentrations, or channel expression levels within the physiologically plausible range. In this HOM regime, spike trains have burst-like irregular firing induced by stochastic switches between attractors. While single-neuron bifurcations leading to HOM dynamics are well-studied (Hesse et al., 2022; Contreras et al., 2021; Hürkey et al., 2023; Schleimer et al., 2021; Niemeyer et al., 2021), their impact on

Over the years, excitatory neurons received much attention for their role in brain dynamics, while the inhibitory neurons were understood to play more of a support, stabilizing role. However, recent advances have revealed that memories are encoded both by excitatory and inhibitory neurons (EI engrams), adding to the evidence of the functional importance of inhibitory neurons beyond mere stability. Still, the computations enabled by such EI engrams are unclear, and the synaptic plasticity rules that could create and sustain them are unknown. Here, we explore in silico a mechanism for EI engram

Synaptic plasticity is key to the brain’s life-long learning capabilities, but the link between changes at individual synapses and emerging network properties remains elusive. Recent meta-learning approaches have started automating the search through the space of possible plasticity rules for plausible candidates. However, these approaches were either limited to narrow sets of rules, or did not investigate any network function beyond stability. Here, we identify thousands of co-active excitatory(E)-to-E, E-to-inhibitory(I), I-to-E and I-to-I Hebbian rules that elicit learning in large recurren

Network models of cortex typically focus on excitatory population dynamics, relegating inhibition to a support role. However, recent studies have shown that memories can be encoded by both excitatory and inhibitory neurons (EI engrams), adding to the evidence of the functional importance of inhibitory neurons beyond mere stabilization. The functional consequence of such EI engrams is unclear, and the synaptic plasticity rules that could sustain them are unknown. Here, we explore in silico a mechanism for EI engrams storage and recall to glean possible computational benefits of such engrams ove

Synaptic plasticity is the cornerstone of learning and memory. However, the mechanistic link between individual synaptic changes and emerging network functions remains elusive. Here, we apply a numerical method for meta-learning plasticity rules---filter Simulation-Based Inference (fSBI)---to in vivo data. We identify thousands of co-active excitatory(E)-to-E, E-to-inhibitory(I), I-to-E and I-to-I rules that can drive learning and memory functions in large recurrent spiking networks. We begin with large search spaces of flexibly-parameterized co-active plasticity rules. We then use fSBI to fil