Computational Neuroscience seminars

Low-dimensional population dynamics can be observed in neural activity recorded from the prefrontal cortex (PFC) of subjects performing simple cognitive tasks. Many studies have shown that recurrent neural networks (RNNs) trained on the same tasks can reproduce qualitatively these state space trajectories, and have used them as models of how neuronal dynamics implement task computations. The PFC is also viewed as a conductor that organizes the communication between cortical areas and provides contextual information. It is then not clear what is its role in solving simple cognitive tasks. Do the low-dimensional trajectories observed in the PFC really correspond to the computations that it performs? Or do they indirectly reflect the computations occurring within the cortical areas projecting to the PFC? To address these questions, we modelled cortical areas with a modular RNN and equipped it with a PFC-like cognitive system. When trained on cognitive tasks, this multi-system brain model can reproduce the low-dimensional population responses observed in neuronal activity as well as classical RNNs. Qualitatively different mechanisms can emerge from the training process when varying some details of the architecture such as the time constants. In particular, there is one class of models where it is the dynamics of the cognitive system that is implementing the task computations, and another where the cognitive system is only necessary to provide contextual information about the task rule as task performance is not impaired when preventing the system from accessing the task inputs. These constitute two different hypotheses about the causal role of the PFC in solving simple cognitive tasks, which could motivate further experiments on the brain.

Spike Timing Dependent Plasticity (STDP) is argued to modulate synaptic strength depending on the timing of pre- and postsynaptic spikes. Physiological experiments identified a variety of temporal kernels: Hebbian, anti-Hebbian and symmetrical LTP/LTD. In this work we present a novel plasticity model, the Voltage-Dependent Pathway Model (VDP), which is able to replicate those distinct kernel types and intermediate versions with varying LTP/LTD ratios and symmetry features. In addition, unlike previous models it retains these characteristics for different neuron models, which allows for comparison of plasticity in different neuron types. The plastic updates depend on the relative strength and activation of separately modeled LTP and LTD pathways, which are modulated by glutamate release and postsynaptic voltage. We used the 15 neuron type parametrizations in the GLIF5 model presented by Teeter et al. (2018) in combination with the VDP to simulate a range of standard plasticity protocols including standard STDP experiments, frequency dependency experiments and low frequency stimulation protocols. Slight variation in kernel stability and frequency effects can be identified between the neuron types, suggesting that the neuron type may have an effect on the effective learning rule. This plasticity model builds a middle ground between biophysical and phenomenological models allowing not just for the combination with more complex and biophysical neuron models, but is also computationally efficient so can be used in network simulations. Therefore it offers the possibility to explore the functional role of the different kernel types and electrophysiological differences in heterogeneous networks in future work.

Decision-making in noisy environments with constant sensory evidence involves integrating sequentially-sampled evidence, a strategy formalized by diffusion models which is supported by decades behavioral and neural findings. By contrast, it is unknown whether this strategy is also used during decision-making when the underlying sensory evidence is expected to change. Here, we trained monkeys to identify the dominant color of a dynamically refreshed checkerboard pattern that doesn't become informative until after a variable delay. Animals' behavioral responses were briefly suppressed after an abrupt change in evidence, and many neurons in the frontal eye field displayed a corresponding dip in activity at this time, similar to the dip frequently observed after stimulus onset. Generalized drift-diffusion models revealed that behavior and neural activity were consistent with a brief suppression of motor output without a change in evidence accumulation itself, in contrast to the popular belief that evidence accumulation is paused or reset. These results suggest that a brief interruption in motor preparation is an important strategy for dealing with changing evidence during perceptual decision making.

The world around us is complex, but at the same time full of meaningful regularities. We can detect, learn and exploit these regularities automatically in an unsupervised manner i.e. without any direct instruction or explicit reward. For example, we effortlessly estimate the average tallness of people in a room, or the boundaries between words in a language. These regularities and prior knowledge, once learned, can affect the way we acquire and interpret new information to build and update our internal model of the world for future decision-making processes. Despite the ubiquity of passively learning from the structured information in the environment, the mechanisms that support learning from real-world experience are largely unknown. By combing sophisticated cognitive tasks in human and rats, neuronal measurements and perturbations in rat and network modelling, we aim to build a multi-level description of how sensory history is utilised in inferring regularities in temporally extended tasks. In this talk, I will specifically focus on a comparative rat and human model, in combination with neural network models to study how past sensory experiences are utilized to impact working memory and decision making behaviours.

It is common for animals to use self-generated movements to actively sense the surrounding environment. For instance, rodents rhythmically move their whiskers to explore the space close to their body. The mouse whisker system has become a standard model to study active sensing and sensorimotor integration through feedback loops. In this work, we developed a bioinspired spiking neural network model of the sensorimotor peripheral whisker system, modelling trigeminal ganglion, trigeminal nuclei, facial nuclei, and central pattern generator neuronal populations. This network was embedded in a virtual mouse robot, exploiting the Neurorobotics Platform, a simulation platform offering a virtual environment to develop and test robots driven by brain-inspired controllers. Eventually, the peripheral whisker system was properly connected to an adaptive cerebellar network controller. The whole system was able to drive active whisking with learning capability, matching neural correlates of behaviour experimentally recorded in mice.

Brains must integrate complex sensory information and compare to past events to generate appropriate behavioral responses. The neural circuit basis of these computations is unclear and the underlying structure unknown. Here, we mapped the comprehensive synaptic wiring diagram of the fruit fly larva brain, which contains 3,013 neurons and 544K synaptic sites. It is the most complete insect connectome to date: 1) Both brain hemispheres are reconstructed, allowing investigation of neural pathways that include contralateral axons, which we found in 37% of brain neurons. 2) All sensory neurons and descending neurons are reconstructed, allowing one to follow signals in an uninterrupted chain—from the sensory periphery, through the brain, to motor neurons in the nerve cord. We developed novel computational tools, allowing us to cluster the brain and investigate how information flows through it. We discovered that feedforward pathways from sensory to descending neurons are multilayered and highly multimodal. Robust feedback was observed at almost all levels of the brain, including descending neurons. We investigated how the brain hemispheres communicate with each other and the nerve cord, leading to identification of novel circuit motifs. This work provides the complete blueprint of a brain and a strong foundation to study the structure-function relationship of neural circuits.

I will build from Niccolo's discussion of the Blockhead argument to argue that having an FeedForward Network (FN) responding like an recurrent network (RN) in a consciousness experiment is not enough to convince us the two are the same with regards to the posession of mental states and conscious experience. I will then argue that a robust functional equivalence between FFN and RN is akso not supported by the mathematical work on the Universal Approximator theorem, and is also unlikely to hold, as a conjecture, given data in cognitive neuroscience; I will argue that an equivalence of RN and FFN may only apply to static functions between input/output layers and not to the temporal patterns or to the network's reactions to structural perturbations. Finally, I review data indicating that consciousness has functional characteristics, such as a flexible control of behavior, and that cognitive/brain dynamics reveal interacting top-down and bottom-up processes, which are necessary for the mediation of such control processes.

Recent advances in optogenetics allow for stimulation of neurons with sub-millisecond spike jitter and single neuron selectivity. Already this precision has revealed new levels of cortical sensitivity: stimulating tens of neurons can yield changes in the mean firing rate of thousands of similarly tuned neurons. This extreme sensitivity suggests that cortical dynamics are near criticality. Criticality is often studied in neural systems as a non-equilibrium thermodynamic process in which scale-free patterns of activity, called avalanches, emerge between distinct states of spontaneous activity. While criticality is well studied, it is still unclear what these distinct states of spontaneous activity are and what responses we expect from stimulation of this activity. By answering these questions, optogenetic stimulation will become a new avenue for approaching criticality and understanding cortical dynamics. Here, for the first time, we study the effects of optogenetic-like stimulation on a model near criticality. We study a model of Inhibitory/Excitatory (I/E) Leaky Integrate and Fire (LIF) spiking neurons which display a region of high sensitivity as seen in experiments. We find that this region of sensitivity is, indeed, near criticality. We derive the Dynamic Mean Field Theory of this model and find that the distinct states of activity are asynchrony and synchrony. We use our theory to characterize response to various types and strengths of optogenetic stimulation. Our model and theory predict that asynchronous, near-critical dynamics can have two qualitatively different responses to stimulation: one characterized by high sensitivity, discrete event responses, and high trial-to-trial variability, and another characterized by low sensitivity, continuous responses with characteristic frequencies, and low trial-to-trial variability. While both response types may be considered near-critical in model space, networks which are closest to criticality show a hybrid of these response effects.

A wide variety of animals and some artificial agents can adapt their behavior to changing cues, contexts, and goals. But what neural network architectures support such behavioral flexibility? Agents with loosely structured network architectures and random connections can be trained over millions of trials to display flexibility in specific tasks, but many animals must adapt and learn with much less experience just to survive. Further, it has been challenging to understand how the structure of trained deep neural networks relates to their functional properties, an important objective for neuroscience. In my talk, I will use a combination of behavioral, physiological and connectomic evidence from the fly to make the case that the built-in modularity and structure of its networks incorporate key aspects of the animal’s ecological niche, enabling rapid flexibility by constraining learning to operate on a restricted parameter set. It is not unlikely that this is also a feature of many biological neural networks across other animals, large and small, and with and without vertebrae.

Neuroscientists have many questions about connectomes that revolve around the ability to compare networks. For example, comparing connectomes could help explain how neural wiring is related to individual differences, genetics, disease, development, or learning. One such question is that of bilateral symmetry: are the left and right sides of a connectome the same? Here, we investigate the bilateral symmetry of a recently presented connectome of an insect brain, the Drosophila larva. We approach this question from the perspective of two-sample testing for networks. First, we show how this question of “sameness” can be framed as a variety of different statistical hypotheses, each with different assumptions. Then, we describe test procedures for each of these hypotheses. We show how these different test procedures perform on both the observed connectome as well as a suite of synthetic perturbations to the connectome. We also point out that these tests require careful attention to parameter alignment and differences in network density in order to provide biologically meaningful results. Taken together, these results provide the first statistical characterization of bilateral symmetry for an entire brain at the single-neuron level, while also giving practical recommendations for future comparisons of connectome networks.

Modeling neural responses to naturalistic stimuli has been instrumental in advancing our understanding of the visual system. Dominant computational modeling efforts in this direction have been deeply rooted in preconceived hypotheses. In contrast, hypothesis-neutral computational methodologies with minimal apriorism which bring neuroscience data directly to bear on the model development process are likely to be much more flexible and effective in modeling and understanding tuning properties throughout the visual system. In this study, we develop a hypothesis-neutral approach and characterize response selectivity in the human visual cortex exhaustively and systematically via response-optimized deep neural network models. First, we leverage the unprecedented scale and quality of the recently released Natural Scenes Dataset to constrain parametrized neural models of higher-order visual systems and achieve novel predictive precision, in some cases, significantly outperforming the predictive success of state-of-the-art task-optimized models. Next, we ask what kinds of functional properties emerge spontaneously in these response-optimized models? We examine trained networks through structural ( feature visualizations) as well as functional analysis (feature verbalizations) by running `virtual' fMRI experiments on large-scale probe datasets. Strikingly, despite no category-level supervision, since the models are solely optimized for brain response prediction from scratch, the units in the networks after optimization act as detectors for semantic concepts like `faces' or `words', thereby providing one of the strongest evidences for categorical selectivity in these visual areas. The observed selectivity in model neurons raises another question: are the category-selective units simply functioning as detectors for their preferred category or are they a by-product of a non-category-specific visual processing mechanism? To investigate this, we create selective deprivations in the visual diet of these response-optimized networks and study semantic selectivity in the resulting `deprived' networks, thereby also shedding light on the role of specific visual experiences in shaping neuronal tuning. Together with this new class of data-driven models and novel model interpretability techniques, our study illustrates that DNN models of visual cortex need not be conceived as obscure models with limited explanatory power, rather as powerful, unifying tools for probing the nature of representations and computations in the brain.

To better understand human scene understanding, we extracted features from images using CLIP, a neural network model of visual concept trained with supervision from natural language. We then constructed voxelwise encoding models to explain whole brain responses arising from viewing natural images from the Natural Scenes Dataset (NSD) - a large-scale fMRI dataset collected at 7T. Our results reveal that CLIP, as compared to convolution based image classification models such as ResNet or AlexNet, as well as language models such as BERT, gives rise to representations that enable better prediction performance - up to a 0.86 correlation with test data and an r-square of 0.75 - in higher-level visual cortex in humans. Moreover, CLIP representations explain distinctly unique variance in these higher-level visual areas as compared to models trained with only images or text. Control experiments show that the improvement in prediction observed with CLIP is not due to architectural differences (transformer vs. convolution) or to the encoding of image captions per se (vs. single object labels). Together our results indicate that CLIP and, more generally, multimodal models trained jointly on images and text, may serve as better candidate models of representation in human higher-level visual cortex. The bridge between language and vision provided by jointly trained models such as CLIP also opens up new and more semantically-rich ways of interpreting the visual brain.

In the deep reinforcement learning (RL) community, motor control problems are usually approached from a reward-based learning perspective. However, humans are often believed to learn motor control through directed error-based learning. Within this learning setting, the control system is assumed to have access to exact error signals and their gradients with respect to the control signal. This is unlike reward-based learning, in which errors are assumed to be unsigned, encoding relative successes and failures. Here, we try to understand the relation between these two approaches, reward- and error- based learning, and ballistic arm reaches. To do so, we test canonical (deep) RL algorithms on a well-known sensorimotor perturbation in neuroscience: mirror-reversal of visual feedback during arm reaching. This test leads us to propose a potentially novel RL algorithm, denoted as model-based deterministic policy gradient (MB-DPG). This RL algorithm draws inspiration from error-based learning to qualitatively reproduce human reaching performance under mirror-reversal. Next, we show MB-DPG outperforms the other canonical (deep) RL algorithms on a single- and a multi- target ballistic reaching task, based on a biomechanical model of the human arm. Finally, we propose MB-DPG may provide an efficient computational framework to help explain error-based learning in neuroscience.

How do humans perceive daily objects of various features and categorize these seemingly intuitive and effortless mental representations? Prior literature focusing on the role of the inferotemporal region (IT) has revealed object category clustering that is consistent with the semantic predefined structure (superordinate, ordinate, subordinate). It has however been debated whether the neural signals in the IT regions are a reflection of such categorical hierarchy [Wen et al.,2018; Bracci et al., 2017]. Visual attributes of images that correlated with semantic and category dimensions may have confounded these prior results. Our study aimed to address this debate by building and comparing models using the DNN AlexNet, to explain the variance in representational dissimilarity matrix (RDM) of neural signals in the IT region. We found that mid and high level perceptual attributes of the DNN model contribute the most to neural RDMs in the IT region. Semantic categories, as in predefined structure, were moderately correlated with mid to high DNN layers (r = [0.24 - 0.36]). Variance partitioning analysis also showed that the IT neural representations were mostly explained by DNN layers, while semantic categorical RDMs brought little additional information. In light of these results, we propose future works should focus more on the specific role IT plays in facilitating the extraction and coding of visual features that lead to the emergence of categorical conceptualizations.

A recent approach to understanding the mammalian visual system is to show correspondence between the sequential stages of processing in the ventral stream with layers in a deep convolutional neural network (DCNN), providing evidence that visual information is processed hierarchically, with successive stages containing ever higher-level information. However, correspondence is usually defined as shared variance between brain region and model layer. We propose that task-relevant variance is a stricter test: If a DCNN layer corresponds to a brain region, then substituting the model’s activity with brain activity should successfully drive the model’s object recognition decision. Using this approach on three datasets (human fMRI and macaque neuron firing rates) we found that in contrast to the hierarchical view, all ventral stream regions corresponded best to later model layers. That is, all regions contain high-level information about object category. We hypothesised that this is due to recurrent connections propagating high-level visual information from later regions back to early regions, in contrast to the exclusively feed-forward connectivity of DCNNs. Using task-relevant correspondence with a late DCNN layer akin to a tracer, we used Granger causal modelling to show late-DCNN correspondence in IT drives correspondence in V4. Our analysis suggests, effectively, that no ventral stream region can be appropriately characterised as ‘early’ beyond 70ms after stimulus presentation, challenging hierarchical models. More broadly, we ask what it means for a model component and brain region to correspond: beyond quantifying shared variance, we must consider the functional role in the computation. We also demonstrate that using a DCNN to decode high-level conceptual information from ventral stream produces a general mapping from brain to model activation space, which generalises to novel classes held-out from training data. This suggests future possibilities for brain-machine interface with high-level conceptual information, beyond current designs that interface with the sensorimotor periphery.

The brain exhibits a rich repertoire of oscillatory patterns organized in space, time and frequency. However, despite ever more-detailed characterizations of spectrally-resolved network patterns, the principles governing oscillatory activity at the system-level remain unclear. Here, we propose that the transient emergence of spatially organized brain rhythms are signatures of weakly stable synchronization between subsets of brain areas, naturally occurring at reduced collective frequencies due to the presence of time delays. To test this mechanism, we build a reduced network model representing interactions between local neuronal populations (with damped oscillatory response at 40Hz) coupled in the human neuroanatomical network. Following theoretical predictions, weakly stable cluster synchronization drives a rich repertoire of short-lived (or metastable) oscillatory modes, whose frequency inversely depends on the number of units, the strength of coupling and the propagation times. Despite the significant degree of reduction, we find a range of model parameters where the frequencies of collective oscillations fall in the range of typical brain rhythms, leading to an optimal fit of the power spectra of magnetoencephalographic signals from 89 heathy individuals. These findings provide a mechanistic scenario for the spontaneous emergence of frequency-specific long-range phase-coupling observed in magneto- and electroencephalographic signals as signatures of resonant modes emerging in the space-time structure of the Connectome, reinforcing the importance of incorporating realistic time delays in network models of oscillatory brain activity.

Inter-areal coherence between cortical field-potentials is a widespread phenomenon and depends on numerous behavioral and cognitive factors. It has been hypothesized that inter-areal coherence reflects phase-synchronization between local oscillations and flexibly gates communication. We reveal an alternative mechanism, where coherence results from and is not the cause of communication, and naturally emerges as a consequence of the fact that spiking activity in a sending area causes post-synaptic inputs both in the same area and in other areas. Consequently, coherence depends in a lawful manner on oscillatory power and phase-locking in a sending area and inter-areal connectivity. We show that changes in oscillatory power explain prominent changes in fronto-parietal beta-coherence with movement and memory, and LGN-V1 gamma-coherence with arousal and visual stimulation. Optogenetic silencing of a receiving area and E/I network simulations demonstrate that afferent synaptic inputs rather than spiking entrainment are the main determinant of inter-areal coherence. These findings suggest that the unique spectral profiles of different brain areas automatically give rise to large-scale inter-areal coherence patterns that follow anatomical connectivity and continuously reconfigure as a function of behavior and cognition.

A myriad of pathological changes associated with epilepsy, including the loss of specific cell types, improper expression of individual ion channels, and synaptic sprouting, can be recast as decreases in cell and circuit heterogeneity. In recent experimental work, we demonstrated that biophysical diversity is a key characteristic of human cortical pyramidal cells, and past theoretical work has shown that neuronal heterogeneity improves a neural circuit’s ability to encode information. Viewed alongside the fact that seizure is an information-poor brain state, these findings motivate the hypothesis that epileptogenesis can be recontextualized as a process where reduction in cellular heterogeneity renders neural circuits less resilient to seizure onset. By comparing whole-cell patch clamp recordings from layer 5 (L5) human cortical pyramidal neurons from epileptogenic and non-epileptogenic tissue, we present the first direct experimental evidence that a significant reduction in neural heterogeneity accompanies epilepsy. We directly implement experimentally-obtained heterogeneity levels in cortical excitatory-inhibitory (E-I) stochastic spiking network models. Low heterogeneity networks display unique dynamics typified by a sudden transition into a hyper-active and synchronous state paralleling ictogenesis. Mean-field analysis reveals a distinct mathematical structure in these networks distinguished by multi-stability. Furthermore, the mathematically characterized linearizing effect of heterogeneity on input-output response functions explains the counter-intuitive experimentally observed reduction in single-cell excitability in epileptogenic neurons. This joint experimental, computational, and mathematical study showcases that decreased neuronal heterogeneity exists in epileptogenic human cortical tissue, that this difference yields dynamical changes in neural networks paralleling ictogenesis, and that there is a fundamental explanation for these dynamics based in mathematically characterized effects of heterogeneity. These interdisciplinary results provide convincing evidence that biophysical diversity imbues neural circuits with resilience to seizure and a new lens through which to view epilepsy, the most common serious neurological disorder in the world, that could reveal new targets for clinical treatment.

Large-scale recordings of neural activity are providing new opportunities to study network-level dynamics with unprecedented detail. However, the sheer volume of data and its dynamical complexity are major barriers to uncovering and interpreting these dynamics. I will present machine learning frameworks that enable inference of dynamics from neuronal population spiking activity on single trials and millisecond timescales, from diverse brain areas, and without regard to behavior. I will then demonstrate extensions that allow recovery of dynamics from two-photon calcium imaging data with surprising precision. Finally, I will discuss our efforts to facilitate comparisons within our field by curating datasets and standardizing model evaluation, including a currently active modeling challenge, the 2021 Neural Latents Benchmark [neurallatents.github.io].