Seminars

We have entered a golden age of artificial intelligence research, driven mainly by the advances in the artificial neural networks over the last several decades. Applications of these techniques—to machine vision, speech recognition, autonomous vehicles, natural language, and many other domains—are coming so quickly that many observers predict that the long-elusive goal of “Artificial General Intelligence” (AGI) is within our grasp. However, we still cannot build a machine capable of building a nest, stalking prey, or loading a dishwasher. I will describe how evolution may have shaped the algorithms that the brain is using to solve some of these challenging problems. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-05-21. Recording duration: 00:44:20.

All animal locomotion is rhythmic,whether it is achieved through undulatory movement of the whole body or the coordination of articulated limbs. Neurobiologists have long studied locomotor circuits that produce rhythmic activity with non-rhythmic input, also called central pattern generators (CPGs). However, the cellular and microcircuit implementation of a walking CPG has not been described for any limbed animal. New comprehensive connectomes of the fruit fly ventral nerve cord (VNC) provide an opportunity to study rhythmogenic walking circuits at a synaptic scale.We use a data-driven network modeling approach to identify and characterize a putative walking CPG in the Drosophila leg motor system.

My talk will consist of three subtopics. The brain remembers episodes not in isolation but with their contextual relationships, such as spatial or temporal proximity. This is an essential feature of the brain’s memory, but the underlying mechanism is yet to be explored. Cell assemblies, or engrams, may provide neural representations for such relationships. First, I will show a class of associative memory models that encode and retrieve multiple memory contents linked by an arbitrary graph structure through experience and demonstrate the crucial role of the balance between two inhibitory subnetwork types in the flexible retrieval of relational memories. Secondly, I propose a theoretical framework to generate a cognitive map, i.e., neural representations of relationships between memory items. This framework aims at the predictive function of the hippocampus and is based on successor representations proposed for reinforcement learning. Intriguingly, the model provides a unified account for grid cells in spatial navigation and concept cells in natural language processing. Finally, I will discuss another crucial role of the hippocampal memory system, memory replay, in a spiking neural network model. Unlike the conventional associative memory models that maintain attractor memory states, this model attempts to maximize the capacity of replayed activity patterns. Our model suggests the crucial role of inhibitory plasticity in optimizing spontaneous memory replay. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-05-14. Recording duration: 00:46:49.

Understanding the neural mechanisms of reward-guided learning is a long-standing goal of computational neuroscience. Recent methodological innovations enable us to collect ever larger neural and behavioral datasets. This presents opportunities to achieve greater understanding of learning in the brain at scale, as well as methodological challenges. In the first part of the talk, I will discuss our recent insights into the mechanisms by which zebra finch songbirds learn to sing. Dopamine has been long thought to guide reward-based trial-and-error learning by encoding reward prediction errors. However, it is unknown whether the learning of natural behaviours, such as developmental vocal learning, occurs through dopamine-based reinforcement. Longitudinal recordings of dopamine and bird songs reveal that dopamine activity is indeed consistent with encoding a reward prediction error during naturalistic learning. In the second part of the talk, I will talk about recent work we are doing at DeepMind to develop tools for automatically discovering interpretable models of behavior directly from animal choice data. Our method, dubbed CogFunSearch, uses LLMs within an evolutionary search process in order to "discover" novel models in the form of Python programs that excel at accurately predicting animal behavior during reward-guided learning. The discovered programs reveal novel patterns of learning and choice behavior that update our understanding of how the brain solves reinforcement learning problems.

The brain is one of the most energy intense organs. Some of this energyis used for neural information processing, however, fruitfly experiments have shown that also learning is metabolically costly. We will present estimates of this cost and introduce a general model of this cost, and compare it to costs in computers. Next, we turn to a supervised artificial network setting and explore a number of strategies that cansave energy need for plasticity. Either by modifying the objective function, by restricting plasticity, or by using less costly transient forms of plasticity. Finally, we will discuss adaptive strategies and possible relevance for biological learning. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-05-07. Recording duration: 00:34:58.

Neural dynamics represent the hard-to-interpret substrate of circuit computations. Advances in large-scale recordings have highlighted the sheer spatiotemporal complexity of circuit dynamics within and across circuits, portraying in detail the difficulty of interpreting such dynamics and relating it to computation. Indeed, even in extremely simplified experimental conditions, one observes high-dimensional temporal dynamics in the relevant circuits. This complexity can be potentially addressed by the notion that not all changes in population activity have equal meaning, i.e., a small change in the evolution of activity along a particular dimension may have a bigger effect on a given computation than a large change in another. We term such conditions dimension-specific computation. Considering motor preparatory activity in a delayed response task we utilized neural recordings performed simultaneously with optogenetic perturbations to probe circuit dynamics. First, we revealed a remarkable robustness in the detailed evolution of certain dimensions of the population activity, beyond what was thought to be the case experimentally and theoretically. Second, the robust dimension in activity space carries nearly all of the decodable behavioral information whereas other non-robust dimensions contained nearly no decodable information, as if the circuit was setup to make informative dimensions stiff, i.e., resistive to perturbations, leaving uninformative dimensions sloppy, i.e., sensitive to perturbations. Third, we show that this robustness can be achieved by a modular organization of circuitry, whereby modules whose dynamics normally evolve independently can correct each other’s dynamics when an individual module is perturbed, a common design feature in robust systems engineering. Finally, we will recent work extending this framework to understanding the neural dynamics underlying preparation of speech.

For decades dynamical systems theory played a pivotal role in theoretical and computational neuroscience, as it links biophysical and biochemical processes to neural computation. In fact, dynamical systems are computationally universal. Rather than hand-crafting computational theories of neural function based on dynamical systems, recent developments in scientific machine learning (ML) and AI suggest that we may be able to infer such dynamical-computational models directly from neurophysiological and behavioral observations. This is called dynamical systems reconstruction (DSR), the learning of generative surrogate models of the underlying dynamics, including its long-term temporal and geometrical properties, from time series data. In my talk I will cover recent ML/AI architectures, training algorithms, and validation procedures for DSR. I will discuss specifically how recent AI architectures for DSR can integrate neuroscience data from multiple modalities (like multiple single-unit recordings and behavioral choices), across diverse time scales, and across many different animals and task designs, into a joint DSR model. This provides first steps toward dynamical systems based AI foundation models for neuroscience. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-04-23. Recording duration: 00:53:24.

Pharmacokinetics in the eye is an important factor for the success of ocular drug delivery and treatment. Pharmacokinetic features determine the feasible routes of drug administration, dosing levels and intervals, and it has impact on eventual drug responses. Several physical, biochemical, and flow-related barriers limit drug exposure of anterior and posterior ocular target tissues during treatment during local (topical, subconjunctival, intravitreal) and systemic administration (intravenous, per oral). Mathematical models integrate joint impact of various barriers on ocular pharmacokinetics (PKs) thereby helping drug development. The models are useful in describing (top-down) and predicting (bottom-up) pharmacokinetics of ocular drugs. This is useful also in the design and development of new drug molecules and drug delivery systems. Furthermore, the models can be used for interspecies translation and probing of disease effects on pharmacokinetics. In this lecture, ocular pharmacokinetics and current modelling methods (noncompartmental analyses, compartmental, physiologically based, and finite element models) are introduced. Future challenges are also highlighted (e.g. intra-tissue distribution, prediction of drug responses, active transport).

Over the past decade, neuroscience, cognitive science and computer science (“AI”) converged to create specific, image-computable, deep neural network models intended to appropriately abstract, emulate and explain the mechanisms of primate ventral visual processing, up to its deepest neural level, the inferior temporal cortex (IT). Because these leading neuroscientific emulation models — aka “digital twins” — are fully observable and machine-executable, they offer predictive and potential application power that our field’s prior conceptual models did not. Our team’s ongoing work is aimed at asking if current digital twin models might support non-invasive, beneficial brain modulation. In this talk, I will describe a key result: we demonstrate that we can use a digital twin to design spatial patterns of light energy that, when “added” to the organism’s retinal input in the context of ongoing natural visual processing, results in precise modulation (i.e. rate bias) of the pattern of a population of IT neurons (where any intended modulation pattern is chosen ahead of time by the scientist). Because the IT visual neural populations are known to directly connect to and modulate downstream neural circuits (e.g. amygdala) that may underlie psychological affective states (e.g. mood and anxiety), this novel basic science may unlock a new, non-invasive application avenue of potential future human clinical benefit. This progress and new impact possibilities resulted from convergent brain science and AI engineering efforts in the domain of visual object intelligence. I will motivate this as just one example of what I believe will unlock in other domains of human intelligence as brain scientists and AI engineers collaborate to develop machine-executable models of the underlying mechanisms of those still-mysterious domains. CARL VAN VREESWIJK MEMORIAL LECTURE 2025. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-04-09. Recording duration: 00:55:49.

The use of electronic circuits to model neural systems goes back to C. Mead and is present in models, from leaky-integrate-and-fire to Hodking-Huxley. Simulating neural network with analog hardware is attractive: it allows to implement neurocomputations in real time without discretization approximations, it has perfect simulation-time scaling with system size, and it provides ready-to-deploy neuromorphic circuit for applications. There are implementations in CMOS technology, however, they are complex, require sophisticated fabrication facilities and, most important, suffer from significant device mismatch. In a radically different approach, based on the concept of memristors, we introduce a neuro-synaptic circuit of unprecedented simplicity, with readily available cheap off-the-shelf electronic components, that can quantitatively reproduce textbook theoretical neuron and synaptic current models. Our neuron circuits can avoid the mismatch problem and are easily tuneable at bio-compatible time-scales. We first introduce a voltage-gated conductance bursting neuron model that produces spike traces that bare striking similarity to experimental recordings. We then introduce synaptic current circuits and show the modularity of our method implementing neurocomputing primitives of basic network motifs, including CPGs. With this "theoretical hardware" approach we show: (i) that neuron adaptation and self-excitation can be viewed as a self-consistent dynamical problem; (ii) that a dynamical memory can be minimally implemented with a single recursive spiking neuron; (iii) that an adaptive membrane current reveals a connection between bursting and the driven harmonic oscillator, perhaps pointing to a neural correlate of the pendular limb motion. Finally we discuss the limitation of the approach to networks of mid-size and its potential application for brain-machine-interfaces, robotics and AI. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-03-19. Recording duration: 00:43:46.

A major challenge in sensory systems is to understand how neurons extract information from the natural environment. Models derived from their responses to artificial stimuli often have a hard time to generalize and predict responses to natural scenes. However, models directly learned on the responses to natural scenes can be hard to interpret. To address this issue, we have recently developed an approach where we add small perturbations to natural scenes and measure how these perturbations change neuronal responses, to better understand the features extracted by sensory neurons. I will show several applications of this approach in the retina, and how it allowed us to uncover non-linear computations performed by ganglion cells, the retinal output. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-03-12. Recording duration: 00:47:06.

How can the hippocampal system transition from episodic one-shot learning to a multi-shot learning regime and what is the utility of the resultant neural representations? This talk will explore the role of the dentate gyrus (DG) anatomy in this context. The canonical DG model suggests it performs pattern separation. More recent experimental results challenge this standard model, suggesting DG function is more complex and also supports the precise binding of objects and events to space and the integration of information across episodes. Very recent studies attribute pattern separation and pattern integration to anatomically distinct parts of the DG (the suprapyramidal blade vs the infrapyramidal blade). We propose a computational model that investigates this distinction. In the model the two processing streams (potentially localized in separate blades) contribute to the storage of distinct episodic memories, and the integration of information across episodes, respectively. The latter forms generalized expectations across episodes, eventually forming a cognitive map. We train the model with two data sets, MNIST and plausible entorhinal cortex inputs. The comparison between the two streams allows for the calculation of a prediction error, which can drive the storage of poorly predicted memories and the forgetting of well-predicted memories. We suggest that differential processing across the DG aids in the iterative construction of spatial cognitive maps to serve the generation of location-dependent expectations, while at the same time preserving episodic memory traces of idiosyncratic events.

Why do individuals, both humans and animals, exhibit personal biases in two-alternative decision-making tasks, even when no clear reason exists to favor one alternative over another? In this talk, I will explore two competing hypotheses to explain these idiosyncratic biases. The first suggests that such tendencies arise from unique personal experiences, where past associations between actions and feedback influence future choices. The second hypothesis proposes that the bias reflects irreducible microscopic heterogeneities in the dynamics of decision-making networks. I will present experimental data and theoretical findings that support the latter hypothesis, shedding new light on the neural mechanisms behind seemingly irrational preferences. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-03-05. Recording duration: 00:50:36.

In the brain, while early sensory areas encode and process external inputs rapidly, higher-association areas are endowed with slow dynamics to benefit information accumulation over time. This property raises the question of why diverse timescales are well localized rather than being mixed up across the cortex, despite high connection density and an abundance of feedback loops that support reliable signal propagation. In this talk, we will address this question by analyzing a large-scale network model of the primate cortex, and we identify a novel dynamical regime termed "interference-free propagation". In this regime, the mean components of the synaptic currents to each downstream area are imbalanced to ensure signals to propagate reliably, while the temporally fluctuating components of the synaptic inputs governed by upstream areas' timescales are largely canceled out, leading to the localization of its own timescale in each downstream area. Our result provides new insights into the operational regime of the cortex, leading to the coexistence of hierarchical timescale localization and reliable signal propagation. Presented in the van Vreeswijk Theoretical Neuroscience Seminar series (formerly WWTNS) on 2025-02-26. Recording duration: 00:48:02.

Brain Emulation Challenge workshop will tackle cutting-edge topics such as ground-truthing for validation, leveraging artificial datasets generated from virtual brain tissue, and the transformative potential of virtual brain platforms, such as applied to the forthcoming Brain Emulation Challenge.

Brain Emulation Challenge workshop will tackle cutting-edge topics such as ground-truthing for validation, leveraging artificial datasets generated from virtual brain tissue, and the transformative potential of virtual brain platforms, such as applied to the forthcoming Brain Emulation Challenge.

Brain Emulation Challenge workshop will tackle cutting-edge topics such as ground-truthing for validation, leveraging artificial datasets generated from virtual brain tissue, and the transformative potential of virtual brain platforms, such as applied to the forthcoming Brain Emulation Challenge.

Brain Emulation Challenge workshop will tackle cutting-edge topics such as ground-truthing for validation, leveraging artificial datasets generated from virtual brain tissue, and the transformative potential of virtual brain platforms, such as applied to the forthcoming Brain Emulation Challenge.

Brain Emulation Challenge workshop will tackle cutting-edge topics such as ground-truthing for validation, leveraging artificial datasets generated from virtual brain tissue, and the transformative potential of virtual brain platforms, such as applied to the forthcoming Brain Emulation Challenge.