Biology subjects

Maass, W.

Publications and source records attributed to Maass, W..

At least 19 recordsLinked to original sources

bioRxiv2026

Sparse cortical dynamics reveal flexible condition-dependent spike-order codes

Cortical networks compute with remarkably sparse spiking activity, yet the circuit mechanisms that organize these few spikes into flexible, condition-dependent temporal codes remain poorly understood. Here we combine analyses of large-scale mouse recordings with a data-driven cortical microcircuit model (CMM) of mouse V1. In recordings, V1 activity exhibits condition-dependent spike-order sequences: peak-latency order varies with task outcome, current image identity, and preceding image identity, while remaining stable under split-half and single-trial analyses. After task optimization by backpropagation through time, the CMM reproduces this sequence-level signature and sparse activity more closely than the matched randomly connected RSNN and rate-RNN controls tested here. Ablations indicate that neuronal heterogeneity and distance-dependent local connectivity each reduce rigid sequential activity, with their combination giving the closest match to measured cortical signatures. Low-dimensional trajectory visualizations and model-silencing experiments further identify high-mutual-information early neurons whose removal perturbs task trajectories and decisions. Together, these results identify a biologically grounded computational principle: neuronal diversity and local connectivity help sparse recurrent networks avoid rigid temporal pipelines and support flexible, condition-dependent spike-order computation, providing candidate design principles for SNNs that exploit flexible temporal codes.

neuroscience↗

bioRxiv2026

Deep-learning-assisted simulation of a cortical circuit: integrating anatomy, physiology and function

Mechanistic understanding of the brain requires models constrained by anatomy, physiology, and functional activity. We present a differentiable simulator and a ~67,000-neuron model of mouse primary visual cortex that integrates multimodal data, including electron-microscopy connectomics, multipatch synaptic physiology, cell-type-resolved intrinsic electrophysiology, and large-scale Neuropixels recordings from diverse cell types. End-to-end training completes on a single GPU in ~6.5 hours while preserving biological constraints. Networks trained only on brief drifting-grating responses reproduce cell-type-specific benchmarks and generalize to new contrasts and natural scenes. We uncover heterogeneous cell-type- and tuning-dependent synaptic organization and show that training preferentially sculpts inhibitory connectivity into distinct cohorts that exert outsized control over network activity. Targeted ablations show that removing biological priors on synaptic weight distributions can preserve functional activity yet disrupt emergent wiring rules. The freely shared models and code facilitate differentiable simulations as a computationally practical framework for studying brain circuit function and mechanisms under biological constraints.

neuroscience↗

bioRxiv2025

Episodic memories enable powerful algorithms for goal-directed decision making that are context-aware and explainable

Most current methods for goal-directed action selection in the face of changing goals and contingencies require DNNs or LLMs. Therefore they are less suited for implementation in edge devices, where low energy-consumption is imperative. The brain shows that similar functionality can be produced with just 20W, even with low precision of synaptic weights. In addition, it endows our action selection with explanations in terms of prior experience, a capability that is missing in those learning approaches where prior experiences are compressed into parameter values. We show that experimental data from neuroscience and cognitive science suggest a concrete alternative where experiences are organized into cognitive maps. It only requires local online learning in shallow networks with binary synaptic weights, and provides explanations for selected actions in terms of related prior experiences. This brain-inspired approach also requires substantially fewer computation steps for navigation in abstract state spaces with deterministic or stochastic actions.

neuroscience↗

bioRxiv2025

Behavioral Time Scale Synaptic Plasticity (BTSP) endows Hyperdimensional Computing with attractor features

Human reasoning depends on reusing pieces of information by binding them together in new ways, thereby "making infinite uses of finite means" (Alexander von Humboldt). Needed for that is a binding mechanism that enables fast composition and decomposition of tokens of information. Binding can easily be implemented in symbolic computations through parentheses and ordering of symbols. But it is a highly nontrivial operation for distributed representations, where the tokens are encoded by activity patterns in large neural networks or large language models, or more abstractly, by a high dimensional vector. Vector Symbolic Architectures (VSAs) provide partial solutions, but are lacking the flexibility of the brain in information retrieval, e.g. retrieving the tokens from a composed representation or retrieval of a composed representation by just providing some tokens as a cue. We show that a mechanism which the brain employs for binding distributed representations, Behavioral Time Scale Synaptic Plasticity (BTSP), overcomes these deficiencies. In particular, it combines binding with attractor features that make information retrieval substantially more flexible and robust. We evaluate its performance on various applications, including encoding and decoding complex visual scenes and hierarchical binding. We also show that it enhances models for natural language processing and abstract brain computation. BTSP-based binding only requires binary synaptic weights and simple local synaptic plasticity, and can therefore easily be implemented through in-memory computing or other innovative designs for energy-efficient AI implementations.

animal behavior and cognition↗

bioRxiv2025

Neural sampling from cognitive maps supportsgoal-directed planning and imagination

AI systems are becoming more intelligent, but at a very high cost in terms of energy consumption and training requirements. In contrast, our brains only require 20W of energy, they learn online, and they can instantly adjust to changing contingencies. This begs the question what data structures, algorithms, and learning methods enable brains to achieve that, and whether these can be ported into artificial devices. We are addressing this question for a core feature of intelligence: The capacity to plan and solve problems, including new problems that involve states which were never encountered before. We examine three tools that brains are likely to employ for achieving that: Cognitive maps, stochastic computing, and compositional coding. We integrate these tools into a transparent neural network model, and demonstrate its power for flexible planning and problem solving. Importantly, this approach is suitable for implementation by in-memory computing and other energy-efficient neuromorphic hardware. In particular, it only requires self-supervised local synaptic plasticity that is suited for on-chip learning. Hence a core feature of brain intelligence, the capacity to generate solutions to problems that were never encountered before, does not require deep neural networks or large language models, and can be implemented in energy-efficient edge devices.

neuroscience↗

bioRxiv2025

A surprising link between cognitive maps, successor-relation based reinforcement learning, and BTSP

Recent recordings from the hippocampus of the human brain suggest that after a few presentations of sequences of unrelated natural images, correlations between emergent neural codes encode the sequential structure of these sequences. We show that this learning process, which is consistent with experimental data on BTSP (Behavioral Time Scale Synaptic Plasticity), creates a cognitive map that enables online-generation of plans for moving to any given goal, both in spatial environments and in abstract graphs. Furthermore, the resulting neural circuits and plasticity rules provide a biologically plausible implementation of Successor Relation based Reinforcement Learning. In addition, this brain-derived approach for learning cognitive maps provides a blueprint for implementing autonomous learning through on-chip plasticity in energy-efficient neuromorphic hardware.

neuroscience↗

bioRxiv2025

A parsimonious model for learning order relations provides a principled explanation of diverse experimental data

Brains are able to extract diverse relations between objects or concepts, and to integrate relational information into cognitive maps that form a basis for decision making. But is has remained a mystery how relational information is learnt and represented by neural networks of the brain. We show that a simple synaptic plasticity rule enables neurons to absorb and represent relational information, and to use it for abstract inference. This inherent capability of neurons and synaptic plasticity is supported by a rigorous mathematical theory. It explains experimental data from the human brain on the emergence of cognitive maps after learning several linear orders, it explains the terminal item effect that enhances transitive inference if a terminal item of an order is involved, and it provides a simple model for fast configuration of internal order representations in the face of new evidence. We also present a rigorous theoretical explanation for the surprising fact that 2D projections of neural representations of linear orders are curved, rather than linear. Since our model does not require stochastic gradient descent in deep neural networks for learning order relations, it is suited for porting the capability to learn multiple relations and using them for fast inference into edge devices with a low energy budget.

neuroscience↗

bioRxiv2023

Data-based large-scale models provide a window into the organization of cortical computations

The neocortex of the brain is one of the most powerful computing devices that exists, but it remains largely open how its computations are organized. Since the neocortex is a 2D tapestry consisting of repeating stereotypical local cortical microcircuits, a key step for solving this problem is to understand how cortical microcircuits compute. We know by now a lot about their connectivity structure and their neuron types, but we are lacking tools for elucidating causal relations between this structure and their computational function. We present a new tool for elucidating this relation: We train large-scale models of cortical microcircuits, which integrate most current knowledge about their structure, for carrying out similar computational tasks as in the brain. We show that the trained model achieves a similar computational performance as the brain, and that it reproduces experimentally found traits of cortical computation and coding that do not appear in neural network models from AI. Furthermore, we reverse-engineer how computations are organized in the model, thereby producing specific hypotheses that can be tested in experimental neuroscience. Altogether we show that cortical microcircuits provide a distinct new neural network paradigm that is of particular interest for neuromorphic engineering because it computes with highly energy-efficient sparse activity. TeaserReverse engineering of cortical computations

neuroscience↗

bioRxiv2023

Memory structure created through behavioral time scale synaptic plasticity

Recent experimental studies in the awake brain have identified a rule for synaptic plasticity that is instrumental for the instantaneous creation of memory traces in area CA1 of the mammalian brain: Behavioral Time scale Synaptic Plasticity. This one-shot learning rule differs in five essential aspects from previously considered plasticity mechanisms. We introduce a transparent model for the core function of this learning rule and establish a theory that enables a principled understanding of the system of memory traces that it creates. Theoretical predictions and numerical simulations show that our model is able to create a functionally powerful content-addressable memory without the need for high-resolution synaptic weights. Furthermore, it reproduces the repulsion effect of human memory, whereby traces for similar memory items are pulled apart to enable differential downstream processing. Altogether, our results create a link between synaptic plasticity in area CA1 of the hippocampus and its network function. They also provide a promising approach for implementing content-addressable memory with on-chip learning capability in highly energy-efficient crossbar arrays of memristors.

neuroscience↗

bioRxiv2023

Competition between bottom-up visual input and internal inhibition generates error neurons in a model of the mouse primary visual cortex

The predictive coding theory, although attractive, is far from being proven. Supporters of this theory agree that bottom-up sensory inputs and top-down predictions of these inputs must be compared in certain types of neurons called error neurons. Excitatory neurons in layer 2/3 (E2/3) of the primary visual cortex (V1) are ideal candidates to act as error neurons, although how these error neurons are generated is poorly understood. In this study, we aimed to gain insight into how the genetically encoded structure of canonical microcircuits in the neocortex implements the emergence of error neurons. To this end, we used a biologically realistic computational model of V1, developed by the Allen Institute, to study the effect that sudden changes in bottom-up input had on the dynamics of E2/3 neurons. We found that the responses of these neurons can be divided into two main classes: one that depolarized (reporting positive errors; dVf neurons) and one that hyperpolarized (reporting negative errors; hVf neurons). Detailed analysis of both network and effective connectivity allowed us to uncover the mechanism that led to the dynamic segregation of these neurons. This mechanism was found to be the competition between the external visual input, originating in the thalamus, and the recurrent inhibition, originating mainly in layers 2/3 and 4. In contrast, we found no evidence of similar division and responses in excitatory infragranular neurons of layers 5 and 6. Our results are in agreement with recent experimental findings and shed light on the mechanisms responsible for the emergence of error neurons.

neuroscience↗

bioRxiv2022

Local prediction-learning enables neural networks to plan

Planning and problem solving are cornerstones of higher brain function. But we do not know how the brain does that. We show that learning of a suitable cognitive map of the problem space suffices. Furthermore, this can be reduced to learning to predict the next observation through local synaptic plasticity. Importantly, the resulting cognitive map encodes relations between actions and observations, and its emergent high-dimensional geometry provides a sense of direction for reaching distant goals. This quasi-Euclidean sense of direction provides a simple heuristic for online planning that works almost as well as the best offline planning algorithms from AI. If the problem space is a physical space, this method automatically extracts structural regularities from the sequence of observations that it receives so that it can generalize to unseen parts. This speeds up learning of navigation in 2D mazes and the locomotion with complex actuator systems, such as legged bodies. The cognitive map learner that we propose does not require a teacher, similar to self-attention networks (Transformers). But in contrast to Transformers, it does not require backpropagation of errors or very large datasets for learning. Hence it provides a blue-print for future energy-efficient neuromorphic hardware that acquires advanced cognitive capabilities through autonomous on-chip learning. One Sentence SummaryLocal synaptic plasticity is all you need for learning to solve problems

neuroscience↗

bioRxiv2021

Analysis of visual processing capabilities and neural coding strategies of a detailed model for laminar cortical microcircuits in mouse V1

Neural networks of the brain that process visual information have structural properties that differ significantly from those of neural networks which are commonly used for visual processing in AI, such as Convolutional Neural Networks (CNNs). But it has remained unknown how these structural differences are related to network function. We analyze visual processing capabilities of a large-scale model for area V1 that arguably provides the most comprehensive accumulation of anatomical and neurophysiological data that is currently available. Its network structure turns out to induce a number of characteristic visual processing capabilities of the brain, in particular the capability to multiplex different visual processing tasks, also on temporally dispersed visual information, with remarkable robustness to noise. This V1 model also exhibits a number of characteristic neural coding properties of the brain, which provide explanations for its superior noise robustness. Since visual processing in the brain is substantially more energy-efficient than implementations of CNNs in common computer hardware, such brain-like neural network models are likely to have also an impact on technology: As blueprints for visual processing in more energy-efficient neuromorphic hardware. TeaserA new generation of neural network models based on neurophysiological data can achieve robust multiplexing capabilities.

neuroscience↗

bioRxiv2021

Analysis of the computational strategy of a detailed laminar cortical microcircuit model for solving the image-change-detection task

The neocortex can be viewed as a tapestry consisting of variations of rather stereotypical local cortical microcircuits. Hence understanding how these microcircuits compute holds the key to understanding brain function. Intense research efforts over several decades have culminated in a detailed model of a generic cortical microcircuit in the primary visual cortex from the Allen Institute. We are presenting here methods and first results for understanding computational properties of this largescale data-based model. We show that it can solve a standard image-change-detection task almost as well as the living brain. Furthermore, we unravel the computational strategy of the model and elucidate the computational role of diverse subtypes of neurons. Altogether this work demonstrates the feasibility and scientific potential of a methodology based on close interaction of detailed data and large-scale computer modelling for understanding brain function.

neuroscience↗

bioRxiv2021

Spike-based symbolic computations on bit strings and numbers

The brain uses recurrent spiking neural networks for higher cognitive functions such as symbolic computations, in particular, mathematical computations. We review the current state of research on spike-based symbolic computations of this type. In addition, we present new results which show that surprisingly small spiking neural networks can perform symbolic computations on bit sequences and numbers and even learn such computations using a biologically plausible learning rule. The resulting networks operate in a rather low firing rate regime, where they could not simply emulate artificial neural networks by encoding continuous values through firing rates. Thus, we propose here a new paradigm for symbolic computation in neural networks that provides concrete hypotheses about the organization of symbolic computations in the brain. The employed spike-based network models are the basis for drastically more energy-efficient computer hardware - neuromorphic hardware. Hence, our results can be seen as creating a bridge from symbolic artificial intelligence to energy-efficient implementation in spike-based neuromorphic hardware.

neuroscience↗

bioRxiv2021

Probabilistic skeletons endow brain-like neural networks with innate computing capabilities

Nature endows networks of spiking neurons in the brain with innate computing capabilities. But it has remained an open problem how the genome achieves that. Experimental data imply that the genome encodes synaptic connection probabilities between neurons depending on their genetic types and spatial distance. We show that this low-dimensional parameterization suffices for programming fundamental computing capabilities into networks of spiking neurons. However, this method is only effective if the network employs a substantial number of different neuron types. This provides an intriguing answer to the open question why the brain employs so many neuron types, many more than were used so far in neural network models. Neural networks whose computational function is induced through their connectivity structure, rather than through synaptic plasticity, are distinguished by short wire length and robustness to weight perturbations. These neural networks features are not only essential for the brain, but also for energy-efficient neuromorphic hardware. Significance statementFundamental computing capabilities of neural networks in the brain are innate, i.e., they do not depend on experience-dependent plasticity. Examples are the capability to recognize odors of poisonous food, and the capability to stand up and walk right after birth. But it has remained unknown how the genetic code can achieve that. A prominent aspect of neural networks of the brain that is under genetic control is the connection probability between neurons of different types. We show that this low-dimensional code suffices for inducing substantial innate computing capabilities in neural networks, provided they have -like the brain- a fair number of different neuron types. Hence under this condition structure can induce computational function in neural networks.

neuroscience↗

bioRxiv2021

Synaptic plasticity in distal dendrites can optimize probabilistic top-down predictions in pyramidal cells

Sensory information is processed by the brain not in a simple feedforward fashion. Rather, bottom-up inputs are combined in pyramidal cells of sensory cortices with top-down information from higher brain areas that arrives through synapses in apical dendrites. The exact functional role of these top-down inputs has remained unknown. A promising abstract model posits that they provide probabilistic priors for bottom-up sensory inputs. We show that this hypothesis is consistent with a large number of experimental about synaptic plasticity in apical dendrites, in particular with the prominent role of NMDA-spikes. We identify conditions under which this synaptic plasticity could approximate the gold standard for self-supervised learning of probabilistic priors: logistic regression. Furthermore, this perspective suggests an additional functional role for the complex structure of the dendritic arborization plays: It enables the neuron to learn substantially more complex landscapes of probabilistic priors.

neuroscience↗

bioRxiv2021

Revisiting the role of synaptic plasticity and network dynamics for fast learning in spiking neural networks

Spike-based neural network models have so far not been able to reproduce the capability of the brain to learn from very few, often even from just a single example. We show that this deficiency of models disappears if one allows synaptic weights to store priors and other information that optimize the learning process, while using the network state to quickly absorb information from new examples. For that, it suffices to include biologically realistic neurons with spike frequency adaptation in the neural network model, and to optimize the learning process through meta-learning. We demonstrate this on a variety of tasks, including fast learning and deletion of attractors, adaptation of motor control to changes in the body, and solving the Morris water maze task - a paradigm for fast learning of navigation to a new goal. Significance StatementIt has often been conjectured that STDP or other rules for synaptic plasticity can only explain some of the learning capabilities of brains. In particular, learning a new task from few trials is likely to engage additional mechanisms. Results from machine learning show that artificial neural networks can learn from few trials by storing information about them in their network state, rather than encoding them in synaptic weights. But these machine learning methods require neural networks with biologically unrealistic LSTM (Long Short Term Memory) units. We show that biologically quite realistic models for neural networks of the brain can exhibit similar capabilities. In particular, these networks are able to store priors that enable learning from very few examples.

neuroscience↗

bioRxiv2020

One-shot learning with spiking neural networks

Understanding how one-shot learning can be accomplished through synaptic plasticity in neural networks of the brain is a major open problem. We propose that approximations to BPTT in recurrent networks of spiking neurons (RSNNs) such as e-prop cannot achieve this because their local synaptic plasticity is gated by learning signals that are rather ad hoc from a biological perspective: Random projections of instantaneously arising losses at the network outputs, analogously as in Broadcast Alignment for feedforward networks. In contrast, synaptic plasticity is gated in the brain by learning signals such as dopamine, which are emitted by specialized brain areas, e.g. VTA. These brain areas have arguably been optimized by evolution to gate synaptic plasticity in such a way that fast learning of survival-relevant tasks is enabled. We found that a corresponding model architecture, where learning signals are emitted by a separate RSNN that is optimized to facilitate fast learning, enables one-shot learning via local synaptic plasticity in RSNNs for large families of learning tasks. The same learning approach also supports fast spike-based learning of posterior probabilities of potential input sources, thereby providing a new basis for probabilistic reasoning in RSNNs. Our new learning approach also solves an open problem in neuromorphic engineering, where on-chip one-shot learning capability is highly desirable for spike-based neuromorphic devices, but could so far not be achieved. Our method can easily be mapped into neuromorphic hardware, and thereby solves this problem.

neuroscience↗