Sander Wessel Keemink
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sander Wessel Keemink has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
- Mark C. W. van Rossum
Affiliations School of Psychology, University of Nottingham, Nottingham, United Kingdom, School of Mathematical Sciences, University of Nottingham, Nottingham, United Kingdom
7 shared publications - Evelyn Dylda
Otto-von-Guericke-University Magdeburg, Institute of Cognitive Neurology and Dementia Research, Magdeburg, Germany
3 shared publications - Janelle M. P. Pakan
Otto von Guericke University
3 shared publications - Michele Nardin
Janelia Research Campus, Howard Hughes Medical Institute, Ashburn, VA, USA
3 shared publications - Nathalie L. Rochefort
Centre for Discovery Brain Sciences, School of Biomedical Sciences, University of Edinburgh, EH8 9XD, Edinburgh, United Kingdom; Simons Initiative for the Developing Brain, University of Edinburgh, EH8 9XD, Edinburgh, United Kingdom
3 shared publications Clemens Boucsein
Bernstein Center Freiburg, Faculty of Biology, University of Freiburg , Freiburg , Germany
2 shared publications
Latest Publications
- Open Access
- 8 October 2026
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (41)
Efficient and robust control with spikes that constrain free energy
Animal brains exhibit remarkable efficiency in perception and action, while being robust to both external and internal perturbations. The means by which brains accomplish this remains, for now, poorly understood, hindering our understanding of animal and human cognition, as well as our own implementation of efficient algorithms for control of dynamical systems. A potential candidate for a robust mechanism of state estimation and action computation is the free energy principle, but existing implementations of this principle have largely relied on conventional, biologically implausible approaches without spikes. We propose an efficient and robust spiking control framework with realistic biological characteristics. The resulting networks function as free energy constrainers, in which neurons only fire if they reduce the free energy of their internal representation. The networks offer efficient operation through highly sparse activity while matching performance with other similar spiking frameworks, and have high resilience against both external (e.g., sensory noise or collisions) and internal perturbations (e.g., synaptic noise and delays or neuron silencing) that such a network would be faced with when deployed by either an organism or an engineer. Overall, our work provides a mathematical account for spiking control through constraining free energy, providing both better insight into how brain networks might leverage their spiking substrate and a route for implementing efficient control algorithms in neuromorphic hardware.
- 8 October 2026
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (41)
Open AccessEfficient and robust control with spikes that constrain free energy
Animal brains exhibit remarkable efficiency in perception and action, while being robust to both external and internal perturbations. The means by which brains accomplish this remains, for now, poorly understood, hindering our understanding of animal and human cognition, as well as our own implementation of efficient algorithms for control of dynamical systems. A potential candidate for a robust mechanism of state estimation and action computation is the free energy principle, but existing implementations of this principle have largely relied on conventional, biologically implausible approaches without spikes. We propose an efficient and robust spiking control framework with realistic biological characteristics. The resulting networks function as free energy constrainers, in which neurons only fire if they reduce the free energy of their internal representation. The networks offer efficient operation through highly sparse activity while matching performance with other similar spiking frameworks, and have high resilience against both external (e.g., sensory noise or collisions) and internal perturbations (e.g., synaptic noise and delays or neuron silencing) that such a network would be faced with when deployed by either an organism or an engineer. Overall, our work provides a mathematical account for spiking control through constraining free energy, providing both better insight into how brain networks might leverage their spiking substrate and a route for implementing efficient control algorithms in neuromorphic hardware.
- Open Access
- 30 July 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (7) , e1014577
Combining sampling and attractor dynamics in spiking models of head direction systems
Neural populations can maintain stable representations of navigation-related variables while integrating uncertain sensory signals. Experimental evidence showed that the precision of head-direction (HD) representations in flies and mice depends on the reliability of sensory cues, highlighting the influence of input uncertainty in attractor-based neural circuits. How do neural dynamics maintain stability while computing under uncertainty? Here, we propose a spiking neural network that unifies two principles — stability through attraction and uncertainty through fluctuation — and reinterpret the HD circuit as an uncertainty-aware integrator rather than a deterministic compass. Specifically, the network uses sampling-based probabilistic inference, where a neural population represents input uncertainty by rapidly fluctuating among likely hypotheses about the world while preserving a stable representation of head direction along an attractor manifold. This formulation suggests why a classical HD “bump” becomes less precise, namely due to rapid fluctuations, reflecting the uncertainty in angular velocity inputs. Our implementation yields experimentally testable predictions: correlated subthreshold voltage fluctuations, multi-timescale nonlinear interaction patterns, and characteristic statistics of bump movement. By combining probabilistic inference with attractor dynamics within one single circuit, our framework suggests how neural populations across species can represent an estimate and its uncertainty through fluctuations while maintaining stability, which could be a general principle for uncertainty-aware computation in noisy biological systems.
- 30 July 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (7) , e1014577
Open AccessCombining sampling and attractor dynamics in spiking models of head direction systems
Neural populations can maintain stable representations of navigation-related variables while integrating uncertain sensory signals. Experimental evidence showed that the precision of head-direction (HD) representations in flies and mice depends on the reliability of sensory cues, highlighting the influence of input uncertainty in attractor-based neural circuits. How do neural dynamics maintain stability while computing under uncertainty? Here, we propose a spiking neural network that unifies two principles — stability through attraction and uncertainty through fluctuation — and reinterpret the HD circuit as an uncertainty-aware integrator rather than a deterministic compass. Specifically, the network uses sampling-based probabilistic inference, where a neural population represents input uncertainty by rapidly fluctuating among likely hypotheses about the world while preserving a stable representation of head direction along an attractor manifold. This formulation suggests why a classical HD “bump” becomes less precise, namely due to rapid fluctuations, reflecting the uncertainty in angular velocity inputs. Our implementation yields experimentally testable predictions: correlated subthreshold voltage fluctuations, multi-timescale nonlinear interaction patterns, and characteristic statistics of bump movement. By combining probabilistic inference with attractor dynamics within one single circuit, our framework suggests how neural populations across species can represent an estimate and its uncertainty through fluctuations while maintaining stability, which could be a general principle for uncertainty-aware computation in noisy biological systems.
- Open Access
- 14 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 603-606
Kozax: Flexible and Scalable Genetic Programming in JAX
- 14 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 603-606
Open AccessKozax: Flexible and Scalable Genetic Programming in JAX
- Open Access
Comparing Brain-Score and ImageNet performance with responses to the scintillating grid illusion
Perceptual illusions are widely used to study brain processing, and are essential for elucidating underlying function. Successful brain models should then also be able to reproduce these illusions. Some of the most successful models for vision are several variants of Deep Neural Networks (DNNs). These models can classify images with human-level accuracy, and many behavioral and activation measurements correlate well with humans and animals. For several networks it was also shown that they can reproduce some human illusions. However, this was typically done for a limited number of networks. In addition, it remains unclear whether the presence of illusions is linked to either how accurate or brain-like the DNNs are. Here, we consider the scintillating grid illusion, to which two DNNs have been shown to respond as if they are impacted by the illusion. We develop a measure for measuring Illusion Strength based on model activation correlations, which takes into account the difference in Illusion Strength between illusion and control images. We then compare the Illusion Strength to both model performance (top-1 ImageNet), and how well the model explains brain activity (Brain-score). We show that the illusion was measurable in a wide variety of networks (41 out of 51). However, we do not find a strong correlation between Illusion Strength and Brain-Score, nor performance. Some models have strong illusion scores but not Brain-Score, or vice-versa, but no model does both well. Finally, this differs strongly between model types, particularly between convolutional and transformer-based architectures, with transformers having low illusion scores. Overall, our work shows that Illusion Strength measures an important metric to consider for assessing brain models, and that some models could still be missing out on some processing important for brain functioning.
Open AccessComparing Brain-Score and ImageNet performance with responses to the scintillating grid illusion
Perceptual illusions are widely used to study brain processing, and are essential for elucidating underlying function. Successful brain models should then also be able to reproduce these illusions. Some of the most successful models for vision are several variants of Deep Neural Networks (DNNs). These models can classify images with human-level accuracy, and many behavioral and activation measurements correlate well with humans and animals. For several networks it was also shown that they can reproduce some human illusions. However, this was typically done for a limited number of networks. In addition, it remains unclear whether the presence of illusions is linked to either how accurate or brain-like the DNNs are. Here, we consider the scintillating grid illusion, to which two DNNs have been shown to respond as if they are impacted by the illusion. We develop a measure for measuring Illusion Strength based on model activation correlations, which takes into account the difference in Illusion Strength between illusion and control images. We then compare the Illusion Strength to both model performance (top-1 ImageNet), and how well the model explains brain activity (Brain-score). We show that the illusion was measurable in a wide variety of networks (41 out of 51). However, we do not find a strong correlation between Illusion Strength and Brain-Score, nor performance. Some models have strong illusion scores but not Brain-Score, or vice-versa, but no model does both well. Finally, this differs strongly between model types, particularly between convolutional and transformer-based architectures, with transformers having low illusion scores. Overall, our work shows that Illusion Strength measures an important metric to consider for assessing brain models, and that some models could still be missing out on some processing important for brain functioning.
- Open Access
- 11 April 2025
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 21 (4) , e1012969
Biases in neural population codes with a few active neurons
Throughout the brain information is coded in the activity of multiple neurons at once, so called population codes. Population codes are a robust and accurate way of coding information. One can evaluate the quality of population coding by trying to read out the code with a decoder, and estimate the encoded stimulus. In particular when neurons are noisy, coding accuracy has extensively been evaluated in terms of the trial-to-trial variation in the estimate. While most decoders yield unbiased estimators if many neurons are actived, when only a few neurons are active, biases readily emerge. That is, even after averaging, a systematic difference between the true stimulus and its estimate remains. We characterize the shape of this bias for different encoding models (rectified cosine tuning and von Mises functions) and show that it can be both attractive or repulsive for different stimulus values. Biases appear for maximum likelihood and Bayesian decoders. The biases have a non-trivial dependence on noise. We also introduce a technique to estimate the bias and variance of Bayesian least square decoders. The work is of interest to those studying neural populations with a few active neurons.
- 11 April 2025
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 21 (4) , e1012969
Open AccessBiases in neural population codes with a few active neurons
Throughout the brain information is coded in the activity of multiple neurons at once, so called population codes. Population codes are a robust and accurate way of coding information. One can evaluate the quality of population coding by trying to read out the code with a decoder, and estimate the encoded stimulus. In particular when neurons are noisy, coding accuracy has extensively been evaluated in terms of the trial-to-trial variation in the estimate. While most decoders yield unbiased estimators if many neurons are actived, when only a few neurons are active, biases readily emerge. That is, even after averaging, a systematic difference between the true stimulus and its estimate remains. We characterize the shape of this bias for different encoding models (rectified cosine tuning and von Mises functions) and show that it can be both attractive or repulsive for different stimulus values. Biases appear for maximum likelihood and Bayesian decoders. The biases have a non-trivial dependence on noise. We also introduce a technique to estimate the bias and variance of Bayesian least square decoders. The work is of interest to those studying neural populations with a few active neurons.
- Open Access
Combining Sampling Methods with Attractor Dynamics in Spiking Models of Head-Direction Systems
Uncertainty is a fundamental aspect of the natural environment, requiring the brain to infer and integrate noisy signals to guide behavior effectively. Sampling-based inference has been proposed as a mechanism for dealing with uncertainty, particularly in early sensory processing. However, it is unclear how to reconcile sampling-based methods with operational principles of higher-order brain areas, such as attractor dynamics of persistent neural representations. In this study, we present a spiking neural network model for the head-direction (HD) system that combines sampling-based inference with attractor dynamics. To achieve this, we derive the required spiking neural network dynamics and interactions to perform sampling from a large family of probability distributions—including variables encoded with Poisson noise. We then propose a method that allows the network to update its estimate of the current head direction by integrating angular velocity samples—derived from noisy inputs—with a pull towards a circular manifold, thereby maintaining consistent attractor dynamics. This model makes specific, testable predictions about the HD system that can be examined in future neurophysiological experiments: it predicts correlated subthreshold voltage fluctuations; distinctive short- and long-term firing correlations among neurons; and characteristic statistics of the movement of the neural activity “bump” representing the head direction. Overall, our approach extends previous theories on probabilistic sampling with spiking neurons, offers a novel perspective on the computations responsible for orientation and navigation, and supports the hypothesis that sampling-based methods can be combined with attractor dynamics to provide a viable framework for studying neural dynamics across the brain.
Open AccessCombining Sampling Methods with Attractor Dynamics in Spiking Models of Head-Direction Systems
Uncertainty is a fundamental aspect of the natural environment, requiring the brain to infer and integrate noisy signals to guide behavior effectively. Sampling-based inference has been proposed as a mechanism for dealing with uncertainty, particularly in early sensory processing. However, it is unclear how to reconcile sampling-based methods with operational principles of higher-order brain areas, such as attractor dynamics of persistent neural representations. In this study, we present a spiking neural network model for the head-direction (HD) system that combines sampling-based inference with attractor dynamics. To achieve this, we derive the required spiking neural network dynamics and interactions to perform sampling from a large family of probability distributions—including variables encoded with Poisson noise. We then propose a method that allows the network to update its estimate of the current head direction by integrating angular velocity samples—derived from noisy inputs—with a pull towards a circular manifold, thereby maintaining consistent attractor dynamics. This model makes specific, testable predictions about the HD system that can be examined in future neurophysiological experiments: it predicts correlated subthreshold voltage fluctuations; distinctive short- and long-term firing correlations among neurons; and characteristic statistics of the movement of the neural activity “bump” representing the head direction. Overall, our approach extends previous theories on probabilistic sampling with spiking neurons, offers a novel perspective on the computations responsible for orientation and navigation, and supports the hypothesis that sampling-based methods can be combined with attractor dynamics to provide a viable framework for studying neural dynamics across the brain.
- Open Access
Biases in population codes with a few active neurons
Throughout the brain information is coded in the activity of multiple neurons at once, so called population codes. Population codes are a robust and accurate way of coding information. One can evaluate the quality of population coding by trying to read out the code with a decoder, and estimate the encoded stimulus. Coding quality has traditionally been evaluated in terms of the trial-to-trial variation in the estimate. However, codes can also display biases. While most decoders yield unbiased estimators in the limit of many active neurons, we find that when only few neurons are active, biases readily emerge for many decoders. We show that the biases turn out to have a non-trivial dependence on noise and tuning curve properties. We also introduce a technique to estimate the bias and variance of Bayesian decoders. Overall, the work expands our understanding of population coding.
Open AccessBiases in population codes with a few active neurons
Throughout the brain information is coded in the activity of multiple neurons at once, so called population codes. Population codes are a robust and accurate way of coding information. One can evaluate the quality of population coding by trying to read out the code with a decoder, and estimate the encoded stimulus. Coding quality has traditionally been evaluated in terms of the trial-to-trial variation in the estimate. However, codes can also display biases. While most decoders yield unbiased estimators in the limit of many active neurons, we find that when only few neurons are active, biases readily emerge for many decoders. We show that the biases turn out to have a non-trivial dependence on noise and tuning curve properties. We also introduce a technique to estimate the bias and variance of Bayesian decoders. Overall, the work expands our understanding of population coding.
- Open Access
Context-dependence of deterministic and nondeterministic contributions to closed-loop steering control
In natural circumstances, sensory systems operate in a closed loop with motor output, whereby actions shape subsequent sensory experiences. A prime example of this is the sensorimotor processing required to align one’s direction of travel, or heading, with one’s goal, a behavior we refer to as steering. In steering, motor outputs work to eliminate errors between the direction of heading and the goal, modifying subsequent errors in the process. The closed-loop nature of the behavior makes it challenging to determine how deterministic and nondeterministic processes contribute to behavior. We overcome this by applying a nonparametric, linear kernel-based analysis to behavioral data of monkeys steering through a virtual environment in two experimental contexts. In a given context, the results were consistent with previous work that described the transformation as a second-order linear system. Classically, the parameters of such second-order models are associated with physical properties of the limb such as viscosity and stiffness that are commonly assumed to be approximately constant. By contrast, we found that the fit kernels differed strongly across tasks in these and other parameters, suggesting context-dependent changes in neural and biomechanical processes. We additionally fit residuals to a simple noise model and found that the form of the noise was highly conserved across both contexts and animals. Strikingly, the fitted noise also closely matched that found previously in a human steering task. Altogether, this work presents a kernel-based analysis that characterizes the context-dependence of deterministic and non-deterministic components of a closed-loop sensorimotor task. New and noteworthy We use nonparametric systems identification techniques to assess the context-dependence of deterministic and nondeterministic contributions to a closed-loop behavior. Classical approaches assume a fixed transformation between sensory input and motor output. Here, we reveal strong changes to the measured sensorimotor transformations with behavioral context. In contrast, noise within the transformation exhibited a consistent form across contexts, subjects, and species. Together, this work demonstrates how context affects the systematic and stochastic components of a closed-loop behavior.
Open AccessContext-dependence of deterministic and nondeterministic contributions to closed-loop steering control
In natural circumstances, sensory systems operate in a closed loop with motor output, whereby actions shape subsequent sensory experiences. A prime example of this is the sensorimotor processing required to align one’s direction of travel, or heading, with one’s goal, a behavior we refer to as steering. In steering, motor outputs work to eliminate errors between the direction of heading and the goal, modifying subsequent errors in the process. The closed-loop nature of the behavior makes it challenging to determine how deterministic and nondeterministic processes contribute to behavior. We overcome this by applying a nonparametric, linear kernel-based analysis to behavioral data of monkeys steering through a virtual environment in two experimental contexts. In a given context, the results were consistent with previous work that described the transformation as a second-order linear system. Classically, the parameters of such second-order models are associated with physical properties of the limb such as viscosity and stiffness that are commonly assumed to be approximately constant. By contrast, we found that the fit kernels differed strongly across tasks in these and other parameters, suggesting context-dependent changes in neural and biomechanical processes. We additionally fit residuals to a simple noise model and found that the form of the noise was highly conserved across both contexts and animals. Strikingly, the fitted noise also closely matched that found previously in a human steering task. Altogether, this work presents a kernel-based analysis that characterizes the context-dependence of deterministic and non-deterministic components of a closed-loop sensorimotor task. New and noteworthy We use nonparametric systems identification techniques to assess the context-dependence of deterministic and nondeterministic contributions to a closed-loop behavior. Classical approaches assume a fixed transformation between sensory input and motor output. Here, we reveal strong changes to the measured sensorimotor transformations with behavioral context. In contrast, noise within the transformation exhibited a consistent form across contexts, subjects, and species. Together, this work demonstrates how context affects the systematic and stochastic components of a closed-loop behavior.
- 15 December 2021
- Published by Peer Community In in Peer Community Journal
- Vol. 1
Nonlinear computations in spiking neural networks through multiplicative synapses
The brain efficiently performs nonlinear computations through its intricate networks of spiking neurons, but how this is done remains elusive. While nonlinear computations can be implemented successfully in spiking neural networks, this requires supervised training and the resulting connectivity can be hard to interpret. In contrast, the required connectivity for any computation in the form of a linear dynamical system can be directly derived and understood with the spike coding network (SCN) framework. These networks also have biologically realistic activity patterns and are highly robust to cell death. Here we extend the SCN framework to directly implement any polynomial dynamical system, without the need for training. This results in networks requiring a mix of synapse types (fast, slow, and multiplicative), which we term multiplicative spike coding networks (mSCNs). Using mSCNs, we demonstrate how to directly derive the required connectivity for several nonlinear dynamical systems. We also show how to 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 demonstrates a novel method for implementing nonlinear computations in spiking neural networks, while keeping the attractive features of standard SCNs (robustness, realistic activity patterns, and interpretable connectivity). Finally, we discuss the biological plausibility of our approach, and how the high accuracy and robustness of the approach may be of interest for neuromorphic computing.
- 15 December 2021
- Published by Peer Community In in Peer Community Journal
- Vol. 1
Nonlinear computations in spiking neural networks through multiplicative synapses
The brain efficiently performs nonlinear computations through its intricate networks of spiking neurons, but how this is done remains elusive. While nonlinear computations can be implemented successfully in spiking neural networks, this requires supervised training and the resulting connectivity can be hard to interpret. In contrast, the required connectivity for any computation in the form of a linear dynamical system can be directly derived and understood with the spike coding network (SCN) framework. These networks also have biologically realistic activity patterns and are highly robust to cell death. Here we extend the SCN framework to directly implement any polynomial dynamical system, without the need for training. This results in networks requiring a mix of synapse types (fast, slow, and multiplicative), which we term multiplicative spike coding networks (mSCNs). Using mSCNs, we demonstrate how to directly derive the required connectivity for several nonlinear dynamical systems. We also show how to 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 demonstrates a novel method for implementing nonlinear computations in spiking neural networks, while keeping the attractive features of standard SCNs (robustness, realistic activity patterns, and interpretable connectivity). Finally, we discuss the biological plausibility of our approach, and how the high accuracy and robustness of the approach may be of interest for neuromorphic computing.
- 1 October 2019
- Published by Elsevier in Current Opinion in Neurobiology
- Vol. 58, 112-121
Decoding and encoding (de)mixed population responses
- 1 October 2019
- Published by Elsevier in Current Opinion in Neurobiology
- Vol. 58, 112-121
Decoding and encoding (de)mixed population responses