Thesis Neural information processing in the fluctuation-driven regime across scales (M/F)

Job Description

Thesis Subject

One of the great mysteries of the awake brain activity is that conscious states are systematically associated with asynchronous brain activity, but how such complex activity may lead to enhanced information processing is still unclear today (Koch et al, 2016). Asynchronous states have also been called fluctuation-driven regimes, because single-neuron firing is determined primarily by the variance of synaptic inputs rather than their mean (Shadlen & Newsome, 1998; van Vreeswijk & Sompolinsky, 1998). This regime fundamentally shapes how information is encoded, transmitted, and modulated, yet its implications for information processing remain unexplored across scales. We will investigate how the fluctuation-driven regime enables responsiveness and information processing from local circuits to whole-brain networks, and how activity-dependent mechanisms dynamically regulate this regime.

At the microscale, balanced AdEx networks are used to characterize how input mean and variance jointly determine information transmission in spiking networks (Brette & Gerstner, 2005). Using mutual information and Fisher information (Brunel & Nadal, 1998), we aim to characterize how variance level optimizes stimulus discriminability and how spike-frequency adaptation dynamically tunes the network toward this optimum. Moderate fluctuations are predicted to maximize information (McDonnell & Ward, 2011), and adaptation to push the network into fluctuation-driven territory when inputs are strong (La Camera et al., 2006).

At the mesoscale, first-order mean-field models of neural populations track only firing rates and cannot capture fluctuation-driven information processing because they ignore single-neuron input variance. El Boustani and Destexhe (2009) developed a second-order mean-field formalism applicable to any neuron model, and subsequent work extended this specifically to AdEx networks (Zerlaut et al., 2018, Depannemaecker et al., 2021). However, these models have focused on dynamical characterization (bifurcations, stability, oscillations) rather than information processing. We aim to build on this existing framework by investigating how variance modulates dynamic range, gain, and mutual information. Furthermore, network structure will be systematically varied, from random to clustered to distance-dependent, to map the structural conditions under which second-order effects become distinguishable.

At the macroscale, the second-order model is embedded into a connectome-based network (Hagmann et al., 2008). This tests whether brain regions in fluctuation-driven states propagate information more efficiently. Using measures of information flow and network responsiveness (Sporns, 2011), we aim to examine whether conscious awake states correspond to globally fluctuation-driven dynamics while unconscious states such as sleep or anesthesia correspond to mean-driven or synchronous regimes (Steriade et al., 1996; Mashour, 2014).

Beyond intrinsic adaptation, we aim to incorporate short-term plasticity and spike-time-dependent plasticity as dynamical rules that reshape synaptic efficacy and network structure over longer timescales. Random networks show negligible covariances, making first- and second-order mean-field models indistinguishable. However, as plasticity strengthens correlations, covariances emerge and the second-order model becomes necessary to capture network dynamics. This asks whether plasticity actively maintains the fluctuation-driven regime that maximizes information processing or pushes the network into a structured covariance-dominated regime. Homeostatic plasticity is also explored as a candidate mechanism for variance homeostasis, keeping single-neuron input variance in a range that optimizes responsivity.

This thesis provides an information-centered account of fluctuation-driven dynamics, linking single-neuron variance, population coding, and whole-brain states to the fundamental question of how neural systems remain responsive.

Your Work Environment

One of the great mysteries of the awake brain activity is that conscious states are systematically associated with asynchronous brain activity, but how such complex activity may lead to enhanced information processing is still unclear today (Koch et al, 2016). Asynchronous states have also been called fluctuation-driven regimes, because single-neuron firing is determined primarily by the variance of synaptic inputs rather than their mean (Shadlen & Newsome, 1998; van Vreeswijk & Sompolinsky, 1998). This regime fundamentally shapes how information is encoded, transmitted, and modulated, yet its implications for information processing remain unexplored across scales. We will investigate how the fluctuation-driven regime enables responsiveness and information processing from local circuits to whole-brain networks, and how activity-dependent mechanisms dynamically regulate this regime.
The proposed position is located at the Institut des Neurosciences Paris-Saclay (NeuroPSI), UMR 9197, which is a Joint Research Unit affiliated with the CNRS and Université Paris-Saclay. The institute is organized into 24 research teams grouped within 4 scientific axes, international scientific networks, 8 technological platforms, and 7 shared services. The institute comprises approximately 260 individuals who develop a multidisciplinary and multi-scale approach to neuroscience. The team consists of 16 members (6 researchers/lecturer-researchers, 5 engineers, and 4 PhD students).

Compensation and benefits

Compensation

2300 € gross monthly

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions