Simulation-based inference

The importance of dendritic structures for neural computation has been recognized since the early days of neuroscience. However, it is only recently, with the increasing availability of data and advances in modelling techniques, that researchers have begun to map out the precise roles of different branching structures and synaptic distributions [1, 2]. Here, we investigate how the nonlinear integration of excitatory and inhibitory currents in dendritic branches can enhance the ability of a neuron to develop input selectivity and E/I balance via Hebbian synaptic plasticity. We model a single ra

The generation of appropriate parameter configurations for high-dimensional hierarchical spiking neural network models is currently one of the slowest and most frustrating tasks in the computational neuroscience workflow. Moreover, many of the currently used methods for multidimensional parametrization only provide single-point estimates, which do not allow deeper insights into the parameter regions of interest and parameter interactions. Having better knowledge about this would allow neuroscientists to understand more about the network dynamics and the relationships between structure and func

Extensive experimental evidence shows that task-relevant neural population dynamics often evolve along trajectories constrained to low-dimensional subspaces [1, 2]. However, how these low-dimensional task representations emerge through learning, and how the neural activity interacts with synaptic plasticity is still an unresolved question. The recent theoretical framework of low-rank recurrent neural networks (lr-RNNs) provides a direct link between connectivity and dynamics by relating structured patterns embedded in the network connectivity to the resulting low-dimensional dynamics [3]. We

In the awake human brain, alpha-band activity (8–13 Hz) is the most prominent electrophysiological signature. Several models exist for alpha-rhythm generation using neural mass models, including the Jansen-Rit, Moran-David-Friston, Robinson-Rennie-Wright, and Liley-Wright models [1]. Fitting these models traditionally relies on manual parameter tuning or computationally expensive parameter searches, and they rarely account for inter-individual variation. However, there is substantial experimental variability in the propensity for robust oscillatory activity in EEG, e.g., influenced by structur

Aims: Social cognition hinges on the fundamental capability to perceive the presence of others. Research indicates that the mere presence of conspecifics can enhance task performance, particularly in simpler tasks, and is closely associated with the activity of context-dependent neural subpopulations. While attentional modulation (attentional-conflict theory) has long been posited as the primary mechanism driving social facilitation, the literature on the neurobiological underpinnings of this modulation remains scant at best. Understanding these mechanisms could yield critical insights into so

…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

Xenopus frog metamorphosis offers a unique opportunity to study motor circuit changes as tadpoles transition from tail-based undulatory swimming to synchronous limb-based movements. Xenopus motor circuits change at a cellular level and generate different motor patterns over development, but the mechanisms driving the changes in motor output during metamorphosis are not well understood.Here, we combine behavioral analysis with computational modeling. We first quantify swimming (tadpole and frog) behavior using the tracking software SLEAP, measuring the decrease in frequency of tail movement, an

…and accessible gradients for parameter inference (i.e., model identification), making it difficult to systematically identify a single configuration that can reproduce experimental data, let alone the full space of data-consistent models and their uncertainty. As a result, SNNs are often reduced a prio

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

Experimental studies have shown that cortical neurons often exhibit a co-tuning of their excitatory and inhibitory receptive fields (i.e. correlation of incoming E/I currents for different input signals). Such co-tuning is hypothesised to be important for efficient computations. Theoretical studies have examined how different plasticity mechanisms can create such co-tuning in feedforward settings, where distinct, uncorrelated inputs to a post-synaptic neuron allow the formation of matching excitatory and inhibitory receptive fields. Still, the cortex is characterized by high levels of recurr