Neuroscience is experiencing unprecedented growth in dataset size both within individual brains and across populations. Large-scale, multimodal datasets are transforming our understanding of brain structure and function, creating opportunities to address previously unexplored questions. However, managing this increasing data volume requires new training and technology approaches. Modern data technologies are reshaping neuroscience by enabling researchers to tackle complex questions within a Ph.D. or postdoctoral timeframe. I will discuss cloud-based platforms such as brainlife.io, that provide scalable, reproducible, and accessible computational infrastructure. Modern data technology can democratize neuroscience, accelerate discovery and foster scientific transparency and collaboration. Concrete examples will illustrate how these technologies can be applied to mapping brain connectivity, studying human learning and development, and developing predictive models for traumatic brain injury (TBI). By integrating cloud computing and scalable data-sharing frameworks, neuroscience can become more impactful, inclusive, and data-driven..

Biological agents continually reconcile the internal states of their brain circuits with incoming sensory and environmental evidence to evaluate when and how to act. The brains of biological agents, including animals and humans, exploit many evolutionary innovations, chiefly modularity—observable at the level of anatomically-defined brain regions, cortical layers, and cell types among others—that can be repurposed in a compositional manner to endow the animal with a highly flexible behavioral repertoire. Accordingly, their behaviors show their own modularity, yet such behavioral modules seldom correspond directly to traditional notions of modularity in brains. It remains unclear how to link neural and behavioral modularity in a compositional manner. We propose a comprehensive framework—compositional modes—to identify overarching compositionality spanning specialized submodules, such as brain regions. Our framework directly links the behavioral repertoire with distributed patterns of population activity, brain-wide, at multiple concurrent spatial and temporal scales. Using whole-brain recordings of zebrafish brains, we introduce an unsupervised pipeline based on neural network models, constrained by biological data, to reveal highly conserved compositional modes across individuals despite the naturalistic (spontaneous or task-independent) nature of their behaviors. These modes provided a scaffolding for other modes that account for the idiosyncratic behavior of each fish. We then demonstrate experimentally that compositional modes can be manipulated in a consistent manner by behavioral and pharmacological perturbations. Our results demonstrate that even natural behavior in different individuals can be decomposed and understood using a relatively small number of neurobehavioral modules—the compositional modes—and elucidate a compositional neural basis of behavior. This approach aligns with recent progress in understanding how reasoning capabilities and internal representational structures develop over the course of learning or training, offering insights into the modularity and flexibility in artificial and biological agents.

Nonstationarity presents a variety of challenges for machine learning systems. One surprising pathology which can arise in nonstationary learning problems is plasticity loss, whereby making progress on new learning objectives becomes more difficult as training progresses. Networks which are unable to adapt in response to changes in their environment experience plateaus or even declines in performance in highly non-stationary domains such as reinforcement learning, where the learner must quickly adapt to new information even after hundreds of millions of optimization steps. The loss of plasticity manifests in a cluster of related empirical phenomena which have been identified by a number of recent works, including the primacy bias, implicit under-parameterization, rank collapse, and capacity loss. While this phenomenon is widely observed, it is still not fully understood. This talk will present exciting recent results which shed light on the mechanisms driving the loss of plasticity in a variety of learning problems and survey methods to maintain network plasticity in non-stationary tasks, with a particular focus on deep reinforcement learning.

Subsecond timing underlies nearly all sensory and motor activities across species and is critical to survival. While subsecond temporal information has been found across cortical and subcortical regions, it is unclear if it is generated locally and intrinsically or if it is a read out of a centralized clock-like mechanism. Indeed, mechanisms of subsecond timing at the circuit level are largely obscure. Primary sensory areas are well-suited to address these question as they have early access to sensory information and provide minimal processing to it: if temporal information is found in these regions, it is likely to be generated intrinsically and locally. We test this hypothesis by training mice to perform an audio-visual temporal pattern sensory discrimination task as we use 2-photon calcium imaging, a technique capable of recording population level activity at single cell resolution, to record activity in primary visual cortex (V1). We have found significant changes in network dynamics through mice’s learning of the task from naive to middle to expert levels. Changes in network dynamics and behavioral performance are well accounted for by an intrinsic model of timing in which the trajectory of q network through high dimensional state space represents temporal sensory information. Conversely, while we found evidence of other temporal encoding models, such as oscillatory activity, we did not find that they accounted for increased performance but were in fact correlated with the intrinsic model itself. These results provide insight into how subsecond temporal information is encoded mechanistically at the circuit level.

Trends in NeuroAI is a reading group hosted by the MedARC Neuroimaging & AI lab (https://medarc.ai/fmri). Title: SwiFT: Swin 4D fMRI Transformer Abstract: Modeling spatiotemporal brain dynamics from high-dimensional data, such as functional Magnetic Resonance Imaging (fMRI), is a formidable task in neuroscience. Existing approaches for fMRI analysis utilize hand-crafted features, but the process of feature extraction risks losing essential information in fMRI scans. To address this challenge, we present SwiFT (Swin 4D fMRI Transformer), a Swin Transformer architecture that can learn brain dynamics directly from fMRI volumes in a memory and computation-efficient manner. SwiFT achieves this by implementing a 4D window multi-head self-attention mechanism and absolute positional embeddings. We evaluate SwiFT using multiple large-scale resting-state fMRI datasets, including the Human Connectome Project (HCP), Adolescent Brain Cognitive Development (ABCD), and UK Biobank (UKB) datasets, to predict sex, age, and cognitive intelligence. Our experimental outcomes reveal that SwiFT consistently outperforms recent state-of-the-art models. Furthermore, by leveraging its end-to-end learning capability, we show that contrastive loss-based self-supervised pre-training of SwiFT can enhance performance on downstream tasks. Additionally, we employ an explainable AI method to identify the brain regions associated with sex classification. To our knowledge, SwiFT is the first Swin Transformer architecture to process dimensional spatiotemporal brain functional data in an end-to-end fashion. Our work holds substantial potential in facilitating scalable learning of functional brain imaging in neuroscience research by reducing the hurdles associated with applying Transformer models to high-dimensional fMRI. Speaker: Junbeom Kwon is a research associate working in Prof. Jiook Cha’s lab at Seoul National University. Paper link: https://arxiv.org/abs/2307.05916

Working memory (WM) is a cognitive function that allows the short-term maintenance and manipulation of information when no longer accessible to the senses. It relies on temporarily storing stimulus features in the activity of neuronal populations. To preserve these dynamics from distraction it has been proposed that pre and post-distraction population activity decomposes into orthogonal subspaces. If orthogonalization is necessary to avoid WM distraction, it should emerge as performance in the task improves. We sought evidence of WM orthogonalization learning and the underlying mechanisms by analyzing calcium imaging data from the prelimbic (PrL) and anterior cingulate (ACC) cortices of mice as they learned to perform an olfactory dual task. The dual task combines an outer Delayed Paired-Association task (DPA) with an inner Go-NoGo task. We examined how neuronal activity reflected the process of protecting the DPA sample information against Go/NoGo distractors. As mice learned the task, we measured the overlap between the neural activity onto the low-dimensional subspaces that encode sample or distractor odors. Early in the training, pre-distraction activity overlapped with both sample and distractor subspaces. Later in the training, pre-distraction activity was strictly confined to the sample subspace, resulting in a more robust sample code. To gain mechanistic insight into how these low-dimensional WM representations evolve with learning we built a recurrent spiking network model of excitatory and inhibitory neurons with low-rank connections. The model links learning to (1) the orthogonalization of sample and distractor WM subspaces and (2) the orthogonalization of each subspace with irrelevant inputs. We validated (1) by measuring the angular distance between the sample and distractor subspaces through learning in the data. Prediction (2) was validated in PrL through the photoinhibition of ACC to PrL inputs, which induced early-training neural dynamics in well-trained animals. In the model, learning drives the network from a double-well attractor toward a more continuous ring attractor regime. We tested signatures for this dynamical evolution in the experimental data by estimating the energy landscape of the dynamics on a one-dimensional ring. In sum, our study defines network dynamics underlying the process of learning to shield WM representations from distracting tasks.

Connor Lane will lead a journal club on the recent BrainLM preprint, a foundation model for fMRI trained using self-supervised masked autoencoder training. Preprint: https://www.biorxiv.org/content/10.1101/2023.09.12.557460v1 Tweeprint: https://twitter.com/david_van_dijk/status/1702336882301112631?t=Q2-U92-BpJUBh9C35iUbUA&s=19

Neural activity is often described in terms of population-level factors extracted from the responses of many neurons. Factors provide a lower-dimensional description with the aim of shedding light on network computations. Yet, mechanistically, computations are performed not by continuously valued factors but by interactions among neurons that spike discretely and variably. Models provide a means of bridging these levels of description. We developed a general method for training model networks of spiking neurons by leveraging factors extracted from either data or firing-rate-based networks. In addition to providing a useful model-building framework, this formalism illustrates how reliable and continuously valued factors can arise from seemingly stochastic spiking. Our framework establishes procedures for embedding this property in network models with different levels of realism. The relationship between spikes and factors in such networks provides a foundation for interpreting (and subtly redefining) commonly used quantities such as firing rates.

Young children have sophisticated representations of their visual and linguistic environment. Where do these representations come from? How much knowledge arises through generic learning mechanisms applied to sensory data, and how much requires more substantive (possibly innate) inductive biases? We examine these questions by training neural networks solely on longitudinal data collected from a single child (Sullivan et al., 2020), consisting of egocentric video and audio streams. Our principal findings are as follows: 1) Based on visual only training, neural networks can acquire high-level visual features that are broadly useful across categorization and segmentation tasks. 2) Based on language only training, networks can acquire meaningful clusters of words and sentence-level syntactic sensitivity. 3) Based on paired visual and language training, networks can acquire word-referent mappings from tens of noisy examples and align their multi-modal conceptual systems. Taken together, our results show how sophisticated visual and linguistic representations can arise through data-driven learning applied to one child’s first-person experience.

Neurons in the rodent hippocampus appear to encode the position of the animal in physical space during movement. Individual ``place cells'' fire in restricted sub-regions of an environment, a feature often taken as evidence that the hippocampus encodes a map of space that subserves navigation. But these same neurons exhibit complex responses to many other variables that defy explanation by position alone, and the hippocampus is known to be more broadly critical for memory formation. Here we elaborate and test a theory of hippocampal coding which produces place cells as a general consequence of efficient memory coding. We constructed neural networks that actively exploit the correlations between memories in order to learn compressed representations of experience. Place cells readily emerged in the trained model, due to the correlations in sensory input between experiences at nearby locations. Notably, these properties were highly sensitive to the compressibility of the sensory environment, with place field size and population coding level in dynamic opposition to optimally encode the correlations between experiences. The effects of learning were also strongly biphasic: nearby locations are represented more similarly following training, while locations with intermediate similarity become increasingly decorrelated, both distance-dependent effects that scaled with the compressibility of the input features. Using virtual reality and 2-photon functional calcium imaging in head-fixed mice, we recorded the simultaneous activity of thousands of hippocampal neurons during virtual exploration to test these predictions. Varying the compressibility of sensory information in the environment produced systematic changes in place cell properties that reflected the changing input statistics, consistent with the theory. We similarly identified representational plasticity during learning, which produced a distance-dependent exchange between compression and pattern separation. These results motivate a more domain-general interpretation of hippocampal computation, one that is naturally compatible with earlier theories on the circuit's importance for episodic memory formation. Work done in collaboration with James Priestley, Lorenzo Posani, Marcus Benna, Attila Losonczy.

In cognitive development, learning more than the input provides is a central challenge. This challenge is especially evident in learning the meaning of numbers. Integers – and the quantities they denote – are potentially infinite, as are the fractional values between every integer. Yet children’s experiences of numbers are necessarily finite. Analogy is a powerful learning mechanism for children to learn novel, abstract concepts from only limited input. However, retrieving proper analogy requires cognitive supports. In this talk, I seek to propose and examine number lines as a mathematical schema of the number system to facilitate both the development of rational number understanding and analogical reasoning. To examine these hypotheses, I will present a series of educational intervention studies with third-to-fifth graders. Results showed that a short, unsupervised intervention of spatial alignment between integers and fractions on number lines produced broad and durable gains in fractional magnitudes. Additionally, training on conceptual knowledge of fractions – that fractions denote magnitude and can be placed on number lines – facilitates explicit analogical reasoning. Together, these studies indicate that analogies can play an important role in rational number learning with the help of number lines as schemas. These studies shed light on helpful practices in STEM education curricula and instructions.

In this presentation, we will discuss adaptations of historical examples of mathematical research to bring out some of the intuitive judgments that accompany the working practice of mathematicians when reasoning by analogy. The main epistemological claim that we will aim to illustrate is that a central part of mathematical training consists in developing a quasi-perceptual capacity to distinguish superficial from deep analogies. We think of this capacity as an instance of Hadamard’s (1954) discriminating faculty of the mathematical mind, whereby one is led to distinguish between mere “hookings” (77) and “relay-results” (80): on the one hand, suggestions or ‘hints’, useful to raise questions but not to back up conjectures; on the other, more significant discoveries, which can be used as an evidentiary source in further mathematical inquiry. In the second part of the presentation, we will present some recent applications of this epistemological framework to mathematics education projects for middle and high schools in Italy.

Can nonhuman animals perceive the relation-between-relations? This intriguing question has been studied over the last 40 years; nonetheless, the extent to which nonhuman species can do so remains controversial. Here, I review empirical evidence suggesting that pigeons, parrots, crows, and baboons join humans in reliably acquiring and transferring relational matching-to-sample (RMTS). Many theorists consider that RMTS captures the essence of analogy, because basic to analogy is appreciating the ‘relation between relations.’ Factors affecting RMTS performance include: prior training experience, the entropy of the sample stimulus, and whether the items that serve as sample stimuli can also serve as choice stimuli.

Understanding Human ability to learn novel concepts from just a few sensory experiences is a fundamental problem in cognitive neuroscience. I will describe a recent work with Ben Sorcher and Surya Ganguli (PNAS, October 2022) in which we propose a simple, biologically plausible, and mathematically tractable neural mechanism for few-shot learning of naturalistic concepts. We posit that the concepts that can be learned from few examples are defined by tightly circumscribed manifolds in the neural firing-rate space of higher-order sensory areas. Discrimination between novel concepts is performed by downstream neurons implementing ‘prototype’ decision rule, in which a test example is classified according to the nearest prototype constructed from the few training examples. We show that prototype few-shot learning achieves high few-shot learning accuracy on natural visual concepts using both macaque inferotemporal cortex representations and deep neural network (DNN) models of these representations. We develop a mathematical theory that links few-shot learning to the geometric properties of the neural concept manifolds and demonstrate its agreement with our numerical simulations across different DNNs as well as different layers. Intriguingly, we observe striking mismatches between the geometry of manifolds in intermediate stages of the primate visual pathway and in trained DNNs. Finally, we show that linguistic descriptors of visual concepts can be used to discriminate images belonging to novel concepts, without any prior visual experience of these concepts (a task known as ‘zero-shot’ learning), indicated a remarkable alignment of manifold representations of concepts in visual and language modalities. I will discuss ongoing effort to extend this work to other high level cognitive tasks.

With recent advances in learning algorithms, recurrent networks of spiking neurons are achieving performance competitive with standard recurrent neural networks. Still, these learning algorithms are limited to small networks of simple spiking neurons and modest-length temporal sequences, as they impose high memory requirements, have difficulty training complex neuron models, and are incompatible with online learning.Taking inspiration from the concept of Liquid Time-Constant (LTCs), we introduce a novel class of spiking neurons, the Liquid Time-Constant Spiking Neuron (LTC-SN), resulting in functionality similar to the gating operation in LSTMs. We integrate these neurons in SNNs that are trained with FPTT and demonstrate that thus trained LTC-SNNs outperform various SNNs trained with BPTT on long sequences while enabling online learning and drastically reducing memory complexity. We show this for several classical benchmarks that can easily be varied in sequence length, like the Add Task and the DVS-gesture benchmark. We also show how FPTT-trained LTC-SNNs can be applied to large convolutional SNNs, where we demonstrate novel state-of-the-art for online learning in SNNs on a number of standard benchmarks (S-MNIST, R-MNIST, DVS-GESTURE) and also show that large feedforward SNNs can be trained successfully in an online manner to near (Fashion-MNIST, DVS-CIFAR10) or exceeding (PS-MNIST, R-MNIST) state-of-the-art performance as obtained with offline BPTT. Finally, the training and memory efficiency of FPTT enables us to directly train SNNs in an end-to-end manner at network sizes and complexity that was previously infeasible: we demonstrate this by training in an end-to-end fashion the first deep and performant spiking neural network for object localization and recognition. Taken together, we out contribution enable for the first time training large-scale complex spiking neural network architectures online and on long temporal sequences.

Biologically plausible spiking neural networks (SNNs) are an emerging architecture for deep learning tasks due to their energy efficiency when implemented on neuromorphic hardware. However, many of the biological features are at best irrelevant and at worst counterproductive when evaluated in the context of task performance and suitability for neuromorphic hardware. In this talk, I will present an alternative paradigm to design deep learning architectures with good task performance in real-world benchmarks while maintaining all the advantages of SNNs. We do this by focusing on two main features – event-based computation and activity sparsity. Starting from the performant gated recurrent unit (GRU) deep learning architecture, we modify it to make it event-based and activity-sparse. The resulting event-based GRU (EGRU) is extremely efficient for both training and inference. At the same time, it achieves performance close to conventional deep learning architectures in challenging tasks such as language modelling, gesture recognition and sequential MNIST.

The brain efficiently performs nonlinear computations through its intricate networks of spiking neurons, but how this is done remains elusive. While recurrent spiking networks implementing linear computations can be directly derived and easily understood (e.g., in the spike coding network (SCN) framework), the connectivity required for nonlinear computations can be harder to interpret, as they require additional non-linearities (e.g., dendritic or synaptic) weighted through supervised training. Here we extend the SCN framework to directly implement any polynomial dynamical system. This results in networks requiring multiplicative synapses, which we term the multiplicative spike coding network (mSCN). We demonstrate how the required connectivity for several nonlinear dynamical systems can be directly derived and implemented in mSCNs, without training. We also show how to precisely carry out higher-order polynomials with coupled networks that use only pair-wise multiplicative synapses, and provide expected numbers of connections for each synapse type. Overall, our work provides an alternative method for implementing nonlinear computations in spiking neural networks, while keeping all the attractive features of standard SCNs such as robustness, irregular and sparse firing, and interpretable connectivity. Finally, we discuss the biological plausibility of mSCNs, and how the high accuracy and robustness of the approach may be of interest for neuromorphic computing.

AI embedded in real systems, such as in satellites, robots and other autonomous devices, must make fast, safe decisions even when the environment changes, or under limitations on the available power; to do so, such systems must be adaptive in real time. To date, edge computing has no real adaptivity – rather the AI must be trained in advance, typically on a large dataset with much computational power needed; once fielded, the AI is frozen: It is unable to use its experience to operate if environment proves outside its training or to improve its expertise; and worse, since datasets cannot cover all possible real-world situations, systems with such frozen intelligent control are likely to fail. Lifelong Learning is the cutting edge of artificial intelligence - encompassing computational methods that allow systems to learn in runtime and incorporate learning for application in new, unanticipated situations. Until recently, this sort of computation has been found exclusively in nature; thus, Lifelong Learning looks to nature, and in particular neuroscience, for its underlying principles and mechanisms and then translates them to this new technology. Our presentation will introduce a number of state-of-the-art approaches to achieve AI adaptive learning, including from the DARPA’s L2M program and subsequent developments. Many environments are affected by temporal changes, such as the time of day, week, season, etc. A way to create adaptive systems which are both small and robust is by making them aware of time and able to comprehend temporal patterns in the environment. We will describe our current research in temporal AI, while also considering power constraints.

How could machines learn as efficiently as humans and animals? How could machines learn to reason and plan? How could machines learn representations of percepts and action plans at multiple levels of abstraction, enabling them to reason, predict, and plan at multiple time horizons? I will propose a possible path towards autonomous intelligent agents, based on a new modular cognitive architecture and a somewhat new self supervised training paradigm. The centerpiece of the proposed architecture is a configurable predictive world model that allows the agent to plan. Behavior and learning are driven by a set of differentiable intrinsic cost functions. The world model uses a new type of energy-based model architecture called H-JEPA (Hierarchical Joint Embedding Predictive Architecture). H-JEPA learns hierarchical abstract representations of the world that are simultaneously maximally informative and maximally predictable.

Results: the main areas which were modified based on participatory feedback were the variety of immersive scenarios to cover a range of ages and interests, the number of levels of complexity to ensure small improvements were measured, the feedback and reward schemes to ensure positive reinforcement, and specific provision for participants with balance issues, who had difficulties when using head-mounted displays. The effectiveness of the finalised BEARS suite will be evaluated in a large-scale clinical trial. We have added in additional login options for other members of the family and based on patient feedback we have improved the accompanying reward schemes. Conclusions: Through participatory design we have developed a training package (BEARS) for young people with bilateral cochlear implants. The training games are appropriate for use by the study population and ultimately should lead to patients taking control of their own management and reducing the reliance upon outpatient-based rehabilitation programmes. Virtual reality training provides a more relevant and engaging approach to rehabilitation for young people.

How could machines learn as efficiently as humans and animals? How could machines learn to reason and plan? How could machines learn representations of percepts and action plans at multiple levels of abstraction, enabling them to reason, predict, and plan at multiple time horizons? I will propose a possible path towards autonomous intelligent agents, based on a new modular cognitive architecture and a somewhat new self-supervised training paradigm. The centerpiece of the proposed architecture is a configurable predictive world model that allows the agent to plan. Behavior and learning are driven by a set of differentiable intrinsic cost functions. The world model uses a new type of energy-based model architecture called H-JEPA (Hierarchical Joint Embedding Predictive Architecture). H-JEPA learns hierarchical abstract representations of the world that are simultaneously maximally informative and maximally predictable. The corresponding working paper is available here:https://openreview.net/forum?id=BZ5a1r-kVsf

Biologically plausible spiking neural networks (SNNs) are an emerging architecture for deep learning tasks due to their energy efficiency when implemented on neuromorphic hardware. However, many of the biological features are at best irrelevant and at worst counterproductive when evaluated in the context of task performance and suitability for neuromorphic hardware. In this talk, I will present an alternative paradigm to design deep learning architectures with good task performance in real-world benchmarks while maintaining all the advantages of SNNs. We do this by focusing on two main features -- event-based computation and activity sparsity. Starting from the performant gated recurrent unit (GRU) deep learning architecture, we modify it to make it event-based and activity-sparse. The resulting event-based GRU (EGRU) is extremely efficient for both training and inference. At the same time, it achieves performance close to conventional deep learning architectures in challenging tasks such as language modelling, gesture recognition and sequential MNIST

New tasks are often similar in structure to old ones. Animals that take advantage of such conserved or “abstract” task structures can master new tasks with minimal training. To understand the neural basis of this abstraction, we developed a novel behavioural paradigm for mice: the “ABCD” task, and recorded from their medial frontal neurons as they learned. Animals learned multiple tasks where they had to visit 4 rewarded locations on a spatial maze in sequence, which defined a sequence of four “task states” (ABCD). Tasks shared the same circular transition structure (… ABCDABCD …) but differed in the spatial arrangement of rewards. As well as improving across tasks, mice inferred that A followed D (i.e. completed the loop) on the very first trial of a new task. This “zero-shot inference” is only possible if animals had learned the abstract structure of the task. Across tasks, individual medial Frontal Cortex (mFC) neurons maintained their tuning to the phase of an animal’s trajectory between rewards but not their tuning to task states, even in the absence of spatial tuning. Intriguingly, groups of mFC neurons formed modules of coherently remapping neurons that maintained their tuning relationships across tasks. Such tuning relationships were expressed as replay/preplay during sleep, consistent with an internal organisation of activity into multiple, task-matched ring attractors. Remarkably, these modules were anchored to spatial locations: neurons were tuned to specific task space “distances” from a particular spatial location. These newly discovered “Spatially Anchored Task clocks” (SATs), suggest a novel algorithm for solving abstraction tasks. Using computational modelling, we show that SATs can perform zero-shot inference on new tasks in the absence of plasticity and guide optimal policy in the absence of continual planning. These findings provide novel insights into the Frontal mechanisms mediating abstraction and flexible behaviour.

Flexible computation is a hallmark of intelligent behavior. Yet, little is known about how neural networks contextually reconfigure for different computations. Humans are able to perform a new task without extensive training, presumably through the composition of elementary processes that were previously learned. Cognitive scientists have long hypothesized the possibility of a compositional neural code, where complex neural computations are made up of constituent components; however, the neural substrate underlying this structure remains elusive in biological and artificial neural networks. Here we identified an algorithmic neural substrate for compositional computation through the study of multitasking artificial recurrent neural networks. Dynamical systems analyses of networks revealed learned computational strategies that mirrored the modular subtask structure of the task-set used for training. Dynamical motifs such as attractors, decision boundaries and rotations were reused across different task computations. For example, tasks that required memory of a continuous circular variable repurposed the same ring attractor. We show that dynamical motifs are implemented by clusters of units and are reused across different contexts, allowing for flexibility and generalization of previously learned computation. Lesioning these clusters resulted in modular effects on network performance: a lesion that destroyed one dynamical motif only minimally perturbed the structure of other dynamical motifs. Finally, modular dynamical motifs could be reconfigured for fast transfer learning. After slow initial learning of dynamical motifs, a subsequent faster stage of learning reconfigured motifs to perform novel tasks. This work contributes to a more fundamental understanding of compositional computation underlying flexible general intelligence in neural systems. We present a conceptual framework that establishes dynamical motifs as a fundamental unit of computation, intermediate between the neuron and the network. As more whole brain imaging studies record neural activity from multiple specialized systems simultaneously, the framework of dynamical motifs will guide questions about specialization and generalization across brain regions.

Spiking recurrent neural networks (RNNs) are a promising tool for solving a wide variety of complex cognitive and motor tasks, due to their rich temporal dynamics and sparse processing. However training spiking RNNs on dedicated neuromorphic hardware is still an open challenge. This is due mainly to the lack of local, hardware-friendly learning mechanisms that can solve the temporal credit assignment problem and ensure stable network dynamics, even when the weight resolution is limited. These challenges are further accentuated, if one resorts to using memristive devices for in-memory computing to resolve the von-Neumann bottleneck problem, at the expense of a substantial increase in variability in both the computation and the working memory of the spiking RNNs. In this talk, I will present our recent work where we introduced a PyTorch simulation framework of memristive crossbar arrays that enables accurate investigation of such challenges. I will show that recently proposed e-prop learning rule can be used to train spiking RNNs whose weights are emulated in the presented simulation framework. Although e-prop locally approximates the ideal synaptic updates, it is difficult to implement the updates on the memristive substrate due to substantial device non-idealities. I will mention several widely adapted weight update schemes that primarily aim to cope with these device non-idealities and demonstrate that accumulating gradients can enable online and efficient training of spiking RNN on memristive substrates.

A core question in human development is how we bring meaning to conventional symbols. This question is deeply connected to understanding how children learn mathematics—a symbol system with unique vocabularies, syntaxes, and written forms. In this talk, I will present findings from a program of research focused on children’s acquisition of place value symbols (i.e., multidigit number meanings). The base-10 symbol system presents a variety of obstacles to children, particularly in English. Children who cannot overcome these obstacles face years of struggle as they progress through the mathematics curriculum of the upper elementary and middle school grades. Through a combination of longitudinal, cross-sectional, and pretest-training-posttest approaches, I aim to illuminate relational learning mechanisms by which children sometimes succeed in mastering the place value system, as well as instructional techniques we might use to help those who do not.

It is widely known that motor learning leads to reorganization changes in the motor cortex. Recently, we have shown that using navigated transcranial magnetic stimulation (TMS) allows us to reliably trace interactions among motor cortical representations (MCRs) of different upper limb muscles. Using this approach, we investigate changes in the MCRs after fine finger movement training. Our preliminary results demonstrated that areas of the APB and ADM and their overlaps tended to increase after finger independence training. Considering the behavioral data, hand dexterity increased for both hands, but the amplitudes of voluntary contraction of the muscles for the APB and ADM did not change significantly. The behavioral results correspond with a previously described suggestion that hand strength and hand dexterity are not directly related as well as an increase in overlaps between MCRs of the trained muscles supports the idea that voluntary muscle relaxation is an active physiological process.

Reservoir computing is a promising framework to study cortical computation, as it is based on continuous, online processing and the requirements and operating principles are compatible with cortical circuit dynamics. However, the framework has issues that limit its scope as a generic model for cortical processing. The most obvious of these is that, in traditional models, learning is restricted to the output projections and takes place in a fully supervised manner. If such an output layer is interpreted at face value as downstream computation, this is biologically questionable. If it is interpreted merely as a demonstration that the network can accurately represent the information, this immediately raises the question of what would be biologically plausible mechanisms for transmitting the information represented by a reservoir and incorporating it in downstream computations. Another major issue is that we have as yet only modest insight into how the structural and dynamical features of a network influence its computational capacity, which is necessary not only for gaining an understanding of those features in biological brains, but also for exploiting reservoir computing as a neuromorphic application. In this talk, I will first demonstrate a method for quantifying the representational capacity of reservoirs without training them on tasks. Based on this technique, which allows systematic comparison of systems, I then present our recent work towards understanding the roles of heterogeneity and connectivity patterns in enhancing both the computational properties of a network and its ability to reliably transmit to downstream networks. Finally, I will give a brief taster of our current efforts to apply the reservoir computing framework to magnetic systems as an approach to neuromorphic computing.

Cognitive and emotional processes are shaped by the dynamic integration of brain and body. A major channel of interoceptive information comes from the heart, where phasic signals are conveyed to the brain to indicate how fast and strong the heart is beating. This talk will discuss how interoceptive processes operate across conscious and unconscious levels to influence emotion and memory. The interoceptive channel is disrupted in distinct ways in individuals with autism and anxiety. Selective interoceptive disturbance is related to symptomatology including dissociation and the transdiagnostic expression of anxiety. Interoceptive training can reduce anxiety, with enhanced interoceptive precision associated with greater insula connectivity following targeted interoceptive feedback. The discrete cardiac effects on emotion and cognition have broad relevance to clinical neuroscience, with implications for peripheral treatment targets and behavioural interventions.

The neural correlates of decision-making have been investigated extensively, and recent work aims to identify under what conditions cortex is actually necessary for making accurate decisions. We discovered that mice with distinct cognitive experiences, beyond sensory and motor learning, use different cortical areas and neural activity patterns to solve the same task, revealing past learning as a critical determinant of whether cortex is necessary for decision tasks. We used optogenetics and calcium imaging to study the necessity and neural activity of multiple cortical areas in mice with different training histories. Posterior parietal cortex and retrosplenial cortex were mostly dispensable for accurate performance of a simple navigation-based visual discrimination task. In contrast, these areas were essential for the same simple task when mice were previously trained on complex tasks with delay periods or association switches. Multi-area calcium imaging showed that, in mice with complex-task experience, single-neuron activity had higher selectivity and neuron-neuron correlations were weaker, leading to codes with higher task information. Therefore, past experience is a key factor in determining whether cortical areas have a causal role in decision tasks.

We have been developing Brain-Computer Interface (BCI) using electrocorticography (ECoG) [1] , which is recorded by electrodes implanted on brain surface, and magnetoencephalography (MEG) [2] , which records the cortical activities non-invasively, for the clinical applications. The invasive BCI using ECoG has been applied for severely paralyzed patient to restore the communication and motor function. The non-invasive BCI using MEG has been applied as a neurofeedback tool to modulate some pathological neural activities to treat some neuropsychiatric disorders. Although these techniques have been developed for clinical application, BCI is also an important tool to investigate neural function. For example, motor BCI records some neural activities in a part of the motor cortex to generate some movements of external devices. Although our motor system consists of complex system including motor cortex, basal ganglia, cerebellum, spinal cord and muscles, the BCI affords us to simplify the motor system with exactly known inputs, outputs and the relation of them. We can investigate the motor system by manipulating the parameters in BCI system. Recently, we are developing some BCIs to visualize and manipulate our perception and mental imagery. Although these BCI has been developed for clinical application, the BCI will be useful to understand our neural system to generate the perception and imagery. In this talk, I will introduce our study of phantom limb pain [3] , that is controlled by MEG-BCI, and the development of a communication BCI using ECoG [4] , that enable the subject to visualize the contents of their mental imagery. And I would like to discuss how much we can control our cortical activities that represent our perception and mental imagery. These examples demonstrate that BCI is a promising tool to visualize and manipulate the perception and imagery and to understand our consciousness. References 1. Yanagisawa, T., Hirata, M., Saitoh, Y., Kishima, H., Matsushita, K., Goto, T., Fukuma, R., Yokoi, H., Kamitani, Y., and Yoshimine, T. (2012). Electrocorticographic control of a prosthetic arm in paralyzed patients. AnnNeurol 71, 353-361. 2. Yanagisawa, T., Fukuma, R., Seymour, B., Hosomi, K., Kishima, H., Shimizu, T., Yokoi, H., Hirata, M., Yoshimine, T., Kamitani, Y., et al. (2016). Induced sensorimotor brain plasticity controls pain in phantom limb patients. Nature communications 7, 13209. 3. Yanagisawa, T., Fukuma, R., Seymour, B., Tanaka, M., Hosomi, K., Yamashita, O., Kishima, H., Kamitani, Y., and Saitoh, Y. (2020). BCI training to move a virtual hand reduces phantom limb pain: A randomized crossover trial. Neurology 95, e417-e426. 4. Ryohei Fukuma, Takufumi Yanagisawa, Shinji Nishimoto, Hidenori Sugano, Kentaro Tamura, Shota Yamamoto, Yasushi Iimura, Yuya Fujita, Satoru Oshino, Naoki Tani, Naoko Koide-Majima, Yukiyasu Kamitani, Haruhiko Kishima (2022). Voluntary control of semantic neural representations by imagery with conflicting visual stimulation. arXiv arXiv:2112.01223.

The SynAGE committee members are thrilled to host ISYNC, the International SynAGE conference on healthy ageing, on 28-30 March 2022 in Magdeburg, Germany. This conference has been entirely organised from young scientists of the SynAGE research training group RTG 2413 (www.synage.de) and represents a unique occasion for researchers from all over the world to bring together and join great talks and sessions with us and our guests. A constantly updated list of our speakers can be found on the conference webpage: www.isync-md.de. During the conference, attendees will have access to a range of symposia which will deal with Glia, Biomarkers and Immunoresponses during ageing to neurodegeneration brain integrity and cognitive function in health and diseases. Moreover, the conference will offer social events especially for young researchers and the possibility to network together in a beautiful and suggestive location where our conference will take place: the Johanniskirche. The event will be happening in person, but due to the current pandemic situation and restrictions we are planning the conference as a hybrid event with lots of technical support to ensure that every participant can follow the talks and take part in the scientific discussions. The registration to our ISYNC conference is free of charge. However, the number of people attending the conference in person is restricted to 100. Afterwards, registrations will be accepted for joining virtually only. The registration is open until 15.02.2022. Especially for PhD and MD Students: Check our available Travel Grants, Poster Prize and SynAGE Award Dinner: https://www.isync-md.de/index.php/phd-md-specials/ If you need any further information don’t hesitate to contact us via email: contact@synage.de. We are looking forward to meet you in 2022 in Magdeburg to discuss about our research and ideas and bless together science. Your ISYNC organization Committee

The brain and its inherent functions can be modified by various forms of learning. Learning-induced changes are seen even in basic perceptual functions. In particular, repeated training in a perceptual task can lead to a significant improvement in the trained task—a phenomenon known as perceptual learning. There has been a long-standing debate about the mechanisms of perceptual learning. In this talk, I will present results from our series of electrophysiological studies. These studies have consistently shown that perceptual learning is mediated by concerted changes in both perceptual and cognitive processes, resulting in improved sensory representation, enhanced top-down influences, and refined readout process.

Analogical reasoning is related to everyday learning and scholastic learning and is a robust predictor of g. Therefore, children's ability to reason by analogy is often measured in a school context to gain insight into children's cognitive and intellectual functioning. Often, the ability to reason by analogy is measured by means of conventional, static instruments. Static tests are criticised by researchers and practitioners to provide an overview of what individuals have learned in the past and for this reason are assumed not to tap into the potential for learning, based on Vygotsky's zone of proximal development. This seminar will focus on children's potential for reasoning by analogy, as measured by means of a dynamic test, which has a test-training-test design. In so doing, the potential relationship between dynamic test outcomes and executive functioning will be explored.

Abstract concepts are a powerful tool to store information efficiently and to make wide-ranging predictions in new situations based on sparse data. Whereas looking-time studies point towards an early emergence of this ability in human infancy, other paradigms like the relational match to sample task often show a failure to detect abstract concepts like same and different until the late preschool years. Similarly, non-human animals have difficulties solving those tasks and often succeed only after long training regimes. Given the huge influence of small task modifications, there is an ongoing debate about the conclusiveness of these findings for the development and phylogenetic distribution of abstract reasoning abilities. Here, we applied the concept of “overhypotheses” which is well known in the infant and cognitive modeling literature to study the capabilities of 3 to 5-year-old children, chimpanzees, and capuchin monkeys in a unified and more ecologically valid task design. In a series of studies, participants themselves sampled reward items from multiple containers or witnessed the sampling process. Only when they detected the abstract pattern governing the reward distributions within and across containers, they could optimally guide their behavior and maximize the reward outcome in a novel test situation. We compared each species’ performance to the predictions of a probabilistic hierarchical Bayesian model capable of forming overhypotheses at a first and second level of abstraction and adapted to their species-specific reward preferences.

Biological brains possess an exceptional ability to infer relevant behavioral responses to a wide range of stimuli from only a few examples. This capacity to generalize beyond the training set has been proven particularly challenging to realize in artificial systems. How neural processes enable this capacity to extrapolate to novel stimuli is a fundamental open question. A prominent but underexplored hypothesis suggests that generalization is facilitated by a low-dimensional organization of collective neural activity, yet evidence for the underlying neural mechanisms remains wanting. Combining network modeling, theory and neural data analysis, we tested this hypothesis in the framework of flexible timing tasks, which rely on the interplay between inputs and recurrent dynamics. We first trained recurrent neural networks on a set of timing tasks while minimizing the dimensionality of neural activity by imposing low-rank constraints on the connectivity, and compared the performance and generalization capabilities with networks trained without any constraint. We then examined the trained networks, characterized the dynamical mechanisms underlying the computations, and verified their predictions in neural recordings. Our key finding is that low-dimensional dynamics strongly increases the ability to extrapolate to inputs outside of the range used in training. Critically, this capacity to generalize relies on controlling the low-dimensional dynamics by a parametric contextual input. We found that this parametric control of extrapolation was based on a mechanism where tonic inputs modulate the dynamics along non-linear manifolds in activity space while preserving their geometry. Comparisons with neural recordings in the dorsomedial frontal cortex of macaque monkeys performing flexible timing tasks confirmed the geometric and dynamical signatures of this mechanism. Altogether, our results tie together a number of previous experimental findings and suggest that the low-dimensional organization of neural dynamics plays a central role in generalizable behaviors.

Making use of visual display technology and human-robotic interfaces, many researchers have illustrated various opportunities to distort visual and physical realities. We have had success with interventions such as error augmentation, sensory crossover, and negative viscosity.  Judicial application of these techniques leads to training situations that enhance the learning process and can restore movement ability after neural injury. I will trace out clinical studies that have employed such technologies to improve the health and function, as well as share some leading-edge insights that include deceiving the patient, moving the "smarts" of software into the hardware, and examining clinical effectiveness

While most experimental tasks aim at isolating simple cognitive processes to study their neural bases, naturalistic behaviour is often complex and multidimensional. I will present two studies revealing previously uncharacterised neural circuits for decision-making in macaques. This was possible thanks to innovative experimental tasks eliciting sophisticated behaviour, bridging the human and non-human primate research traditions. Firstly, I will describe a specialised medial frontal circuit for novel choice in macaques. Traditionally, monkeys receive extensive training before neural data can be acquired, while a hallmark of human cognition is the ability to act in novel situations. I will show how this medial frontal circuit can combine the values of multiple attributes for each available novel item on-the-fly to enable efficient novel choices. This integration process is associated with a hexagonal symmetry pattern in the BOLD response, consistent with a grid-like representation of the space of all available options. We prove the causal role played by this circuit by showing that focussed transcranial ultrasound neuromodulation impairs optimal choice based on attribute integration and forces the subjects to default to a simpler heuristic decision strategy. Secondly, I will present an ongoing project addressing the neural mechanisms driving behaviour shifts during an evidence accumulation task that requires subjects to trade speed for accuracy. While perceptual decision-making in general has been thoroughly studied, both cognitively and neurally, the reasons why speed and/or accuracy are adjusted, and the associated neural mechanisms, have received little attention. We describe two orthogonal dimensions in which behaviour can vary (traditional speed-accuracy trade-off and efficiency) and we uncover independent neural circuits concerned with changes in strategy and fluctuations in the engagement level. The former involves the frontopolar cortex, while the latter is associated with the insula and a network of subcortical structures including the habenula.

Object recognition relies on the complex visual representations in cortical areas at the top of the ventral stream hierarchy. While these are thought to be derived from low-level stages of visual processing, this has not been shown, yet. Here, I describe the results of two projects exploring the contributions of primary visual cortex (V1) processing to object recognition using artificial neural networks (ANNs). First, we developed hundreds of ANN-based V1 models and evaluated how their single neurons approximate those in the macaque V1. We found that, for some models, single neurons in intermediate layers are similar to their biological counterparts, and that the distributions of their response properties approximately match those in V1. Furthermore, we observed that models that better matched macaque V1 were also more aligned with human behavior, suggesting that object recognition is derived from low-level. Motivated by these results, we then studied how an ANN’s robustness to image perturbations relates to its ability to predict V1 responses. Despite their high performance in object recognition tasks, ANNs can be fooled by imperceptibly small, explicitly crafted perturbations. We observed that ANNs that better predicted V1 neuronal activity were also more robust to adversarial attacks. Inspired by this, we developed VOneNets, a new class of hybrid ANN vision models. Each VOneNet contains a fixed neural network front-end that simulates primate V1 followed by a neural network back-end adapted from current computer vision models. After training, VOneNets were substantially more robust, outperforming state-of-the-art methods on a set of perturbations. While current neural network architectures are arguably brain-inspired, these results demonstrate that more precisely mimicking just one stage of the primate visual system leads to new gains in computer vision applications and results in better models of the primate ventral stream and object recognition behavior.

Predictive coding represents a promising framework for understanding brain function, postulating that the brain continuously inhibits predictable sensory input, ensuring a preferential processing of surprising elements. A central aspect of this view on cortical computation is its hierarchical connectivity, involving recurrent message passing between excitatory bottom-up signals and inhibitory top-down feedback. Here we use computational modelling to demonstrate that such architectural hard-wiring is not necessary. Rather, predictive coding is shown to emerge as a consequence of energy efficiency, a fundamental requirement of neural processing. When training recurrent neural networks to minimise their energy consumption while operating in predictive environments, the networks self-organise into prediction and error units with appropriate inhibitory and excitatory interconnections and learn to inhibit predictable sensory input. We demonstrate that prediction units can reliably be identified through biases in their median preactivation, pointing towards a fundamental property of prediction units in the predictive coding framework. Moving beyond the view of purely top-down driven predictions, we demonstrate via virtual lesioning experiments that networks perform predictions on two timescales: fast lateral predictions among sensory units and slower prediction cycles that integrate evidence over time. Our results, which replicate across two separate data sets, suggest that predictive coding can be interpreted as a natural consequence of energy efficiency. More generally, they raise the question which other computational principles of brain function can be understood as a result of physical constraints posed by the brain, opening up a new area of bio-inspired, machine learning-powered neuroscience research.

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.

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.

For the first time, we are holding a NMC around the globe session, a panel of computational neuroscientists working in different continents who are willing to discuss their challenges and milestones in doing science and training researchers in their home country. We hope that our panelists can share their barriers, what they define as accomplishments and how they would like the future of computational neuroscience to evolve locally and internationally with our diverse NMC audience.

It is well established that the opportunity to compare learning stimuli in a novel word learning/extension task elicits a larger number of conceptually relevant generalizations than standard no-comparison conditions. I will present results suggesting that the effectiveness of comparison depends on factors such as semantic distance, number of training items, dimension distinctiveness and interactions with age. I will address these issues in the case of familiar and unfamiliar object nouns and relational nouns. The alignment strategies followed by children during learning and at test (i.e., when learning items are compared and how children reach a solution) will be described with eye-tracking data. We will also assess the extent to which children’s performance in these tasks are associated with executive functions (inhibition and flexibility) and world knowledge. Finally, we will consider these issues in children with cognitive deficits (Intellectual deficiency, DLD)

In the past few years, there has been growing evidence that the basic ability for relational generalization starts in early infancy, with 3-month-olds seeming to learn relational abstractions with little training. Further, work with toddlers seem to suggest that relational generalizations are no more difficult than those based on objects, and they can readily consider both simultaneously. Likewise, causal learning research with adults suggests that people infer causal relationships at multiple levels of abstraction simultaneously as they learn about novel causal systems. These findings all appear counter to theories of concept learning that posit when concepts are first learned they tend to be concrete (tied to specific contexts and features) and abstraction proceeds incrementally as learners encounter more examples. The current talk will not question the veracity of any of these findings but will present several others from my and others’ research on relational learning that suggests that when the perceptual or conceptual content becomes more complex, patterns of incremental abstraction re-emerge. Further, the specific contexts and task parameters that support or hinder abstraction reveal the underlying cognitive processes. I will then consider whether the models that posit simultaneous, immediate learning at multiple levels of abstraction can accommodate these more complex patterns.

Many patients presenting to neurology either have primary sleep disorders or suffer from sleep comorbidity. Knowledge on the diagnosis, differential diagnostic considerations, and management of these disorders is therefore mandatory for the general neurologist. This comprehensive course may serve to fulfill part of the preparation requirements for trainees seeking to complete the Royal College Examinations in Neurology. This training course is for R4 and R5 residents in Canadian neurology training programs as well as neurologists.