Associative emotional learning enables organisms to adaptively link pleasant or unpleasant outcomes to the presence of predictive stimuli. Whereas computational models such as the Rescorla-Wagner model have shed light on this important function, the limitations of these models are also known, especially when they are applied to neural data. The advent of deep neural networks has opened another avenue for modeling associative emotional learning. In this work, we proposed a deep neural network model of visual valence processing, consisting of a visual module that encodes complex natural scenes and a module that recognizes their emotional significance in terms of valence, a key dimension of emotion, and tested a novel Pavlovian learning paradigm on the model. The results showed that with learning, the model reproduced several observations from human associative learning studies, including association formation and generalization, and that the neural representations of the conditioned and the unconditioned stimuli became increasingly aligned at both the single unit and the neural population level. Comparison between the model and human experimental data provided further validation of our approach. This study thus suggests that deep neural network models, when combined with appropriate learning algorithms, can be used to model behavioral and neural signatures of associative emotion/valence learning.

During the past decades, hippocampal formation has undergone extensive studies, leading researchers to identify a vast collection of cells with functional properties. Several investigations, supported by carefully crafted models, have examined the origin of such cells. The most recent models hypothesize that temporal sequences underlie the observed spatial properties. We aim at investigating whether a random recurrent structure is sufficient to allow such latent sequence to appear. To do so, we simulated an agent with egocentric sensory inputs that must navigate and alternate choices at intersections. We were subsequently able to identify several splitter cells inside the model. Remarkably, when we systematically lesioned the identified splitter cells, the model's behavioral performance remained intact: in the vast majority of cases, new splitter cells reemerged through network reorganization, while in the remaining cases, the task was solved without any detectable splitter cells, demonstrating that splitter cells are not necessary to the task resolution. Position, orientation, and decision representations could also be successfully decoded from the reservoir activity, even after repeated lesioning. Subspace alignment analysis further revealed that this reorganization preserves the task-relevant population geometry while redistributing activity within the null subspace, with the trajectory-encoding dimension rotating in neuron space across lesions. Together, these findings demonstrate that splitter cell activity is primarily task driven and does not derive from a specific architecture or learning rule: splitter cells emerge generically across random recurrent networks that successfully solve the task, across a broad and robust range of dynamical parameters and are not necessary for task performance. Our results therefore challenge the notion of functional necessity for specific neural populations.

This article presents a cortically motivated framework for geometry-aware sparse reservoir computing with efficient online readout adaptation. The architecture is based on FitzHugh-Nagumo units coupled through sparse connectivity motifs informed by cortical statistics (reciprocity, triadic closure, and Dale's law), resulting in a structured and biologically plausible recurrent matrix. The readout weights are trained by means of an anchor-proximal ridge objective applied to streaming time windows, ensuring temporally stable online learning with a low memory footprint. It presents the results of numerical experiments on a temporal decision-making benchmark and an image-based long-range segmentation task. The proposed sparse reservoir, combined with the windowed proximal readout, is shown to achieve competitive accuracy while substantially reducing recurrent and state-storage requirements. A favorable trade-off of biological plausibility, predictive performance, and computational efficiency is thereby demonstrated.

Recordings of increasingly large neural populations have revealed that the firing of individual neurons is highly coordinated. When viewed in the space of all possible patterns, the collective activity forms nonlinear structures called neural manifolds. Because such structures are observed even at rest or during sleep, an important hypothesis is that activity manifolds may correspond to continuous attractors shaped by recurrent connectivity between neurons. Classical models of recurrent networks have shown that continuous attractors can be generated by specific symmetries in the connectivity. Although a variety of attractor network models have been studied, general principles linking network connectivity and the geometry of attractors remain to be formulated. Here, we address this question by using group representation theory to formalize the relationship between the symmetries in recurrent connectivity and the resulting fixed-point manifolds. We start by revisiting the classical ring model, a continuous attractor network generating a circular manifold. Interpreting its connectivity as a circular convolution, we draw a parallel with feedforward convolutional neural networks. Building on principles of geometric deep learning, we then generalize this architecture to a broad range of symmetries using group representation theory. Specifically, we introduce a new class of equivariant recurrent neural networks, where the connectivity is based on group convolution. Using the group Fourier transform, we reduce such networks to low-rank models. This formulation gives us a low-dimensional nonlinear description that can be fully analyzed to determine the symmetry, dimensionality, and stability of fixed-point manifolds. Our results underline the importance of stability considerations: for a connectivity with a given symmetry, depending on parameters, several manifolds with different topology can coexist, some stable and others consisting of saddle points. Our framework unifies a variety of existing models and offers a principled approach to build new types of continuous attractor models of neural systems.

Depolarizing current injections produced a rhythmic bursting of action potentials, a bursting oscillation, in a set of local interneurons in the lateral geniculate nucleus (LGN) of rats. The current dynamics underlying this firing pattern have not been determined, though this cell type constitutes an important cellular component of thalamocortical circuitry and contributes to both pathologic and nonpathologic brain states. We thus investigated the source of the bursting oscillation using pharmacological manipulations in LGN slices in vitro and in silico. We found that selective blockade of calcium channel subtypes revealed that high-threshold calcium currents IL and IP contributed strongly to the oscillation. In addition, increased extracellular K+ concentration (decreased K+ currents) eliminated the oscillation. In addition, selective blockade of K+ channel subtypes demonstrated that the calcium-sensitive potassium current (IAHP) was of primary importance. A morphologically simplified, multicompartment model of the thalamic interneuron characterized the oscillation as follows, First, the low-threshold calcium current (IT) provided the strong initial burst characteristic of the oscillation. In addition, alternating fluxes through high-threshold calcium channels and IAHP then provided the continuing oscillation's burst and interburst periods, respectively. This interplay between IL and IAHP contrasts with the current dynamics underlying oscillations in thalamocortical and reticularis neurons, which primarily involve IT and hyperpolarization-activated cation current IH, or IT and IAHP, respectively. These findings thus point to a novel electrophysiological mechanism for generating intrinsic oscillations in a major thalamic cell type. Because local interneurons can sculpt the behavior of thalamocortical circuits, these results suggest new targets for the manipulation of ascending thalamocortical network activity.

Vernon Mountcastle hypothesized that the basis for intelligence in mammals is the replication of a general computational unit, the cortical column. The Thousand Brains Theory proposed that each column is a sensorimotor system, capable of learning structured models of objects by integrating sensory input over multiple movements. Previous papers on the Thousand Brains Theory focused on the computations that occur within individual cortical columns and how columns can use long-range connections to rapidly reach a consensus. However, several prominent long-range connection types in the neocortex were not addressed by the theory. These include hierarchical feedforward and feedback connections, as well as those that go through the thalamus. In addition, several theoretical requirements were not addressed. These include how the cortex learns compositional objects and how information is converted from the egocentric perspective of sensors to the allocentric perspective of models in the cortex. In this letter, we extend the Thousand Brains Theory to address these issues. We begin by reviewing the anatomy of long-range neocortical connections, arguing that they form a heterarchy, rather than hierarchy, which has made their functions challenging to understand through existing theoretical models. We then propose specific roles for each of these connections. First, the thalamus converts the orientation of features and movement information from an egocentric perspective to the allocentric perspective of learned models. Second, hierarchical feedforward, feedback, and cortico-thalamo-cortical projections enable columns to learn compositional models. We discuss the relationship of our proposals to experimental findings at the levels of anatomy, neurophysiology, and behavior, along with testable predictions for future experimental work. We conclude by discussing the implications of the extended Thousand Brains Theory for the design of artificial intelligence.

A key challenge in biophysically detailed modeling is determining functionally effective synaptic weights. Here, we introduce a dendritic voltage surrogate that couples passive dendritic transfer impedances with the exact active currents computed from the original model, yielding explicit, differentiable approximations of membrane voltage. The resulting gradients can achieve effective synaptic learning in active dendrites under transient conditions. In detailed pyramidal neuron models with somatic bursting and dendritic calcium plateau, the surrogate enables reconstruction of subthreshold and bursting membrane potentials and supports supervised fitting of target responses. At network scale, we train a detailed 136-neuron model of Caenorhabditis elegans to match an experimentally measured whole-brain correlation matrix using a multi-GPU framework. Our approach links interpretable dendritic computation with scalable optimization, providing an effective synaptic learning framework for detailed neurons and networks.

There is a general consensus in theories of human speech recognition that humans engage in predictive processing during online speech processing. There are also claims that predictive processing is indicative of the operation of a predictive coding (PC) mechanism. PC is a generative, hierarchical feedback framework where feedback signals consist of input predictions, while feedforward signals consist primarily of prediction errors (PE). Some researchers have taken decreased neural signals when inputs conform to expectations as evidence for PC and claim that other possible explanatory frameworks (e.g., interactive activation) are incompatible with reported reductions in PE. However, these claims have been advanced using narrow-scope computational implementations of PC without known abilities to adequately (i.e., plausibly) simulate broader human spoken word recognition (SWR) behavioral phenomena. Here, we present the first known neurally and behaviorally adequate, mathematically formal neural network PC model of time-dependent human SWR. After demonstrating that the new model is able to emulate a fundamental, empirical signature of human SWR behavior (time course of lexical activation and competition), we compare model dynamics to neural performance targets that have been touted as hallmarks of PC in SWR. While the new model readily exhibits predictive processing (anticipatory activation of phonemes consistent with lexical knowledge) and reduced neural activity when inputs match expectations (a necessary component of PC) more subtle patterns of neural activity under priming and noise conditions, previously proposed as diagnostic of PC do not emerge appreciably. This suggests that such patterns may not be hallmarks of PC. We discuss implications for PC-based theories of SWR.

Sensory pathways differ widely across species and modalities in anatomy and function. Yet many solve variants of the same ancient, pervasive problem: quickly detecting meaningful changes in the world. Sometimes animals need to make these detections from noisy signals with millisecond precision to survive. We suggest how a single spiking neuron can address this selection pressure for quick detection. We present a simple relationship linking neural voltage dynamics to online hypothesis testing that we call the time constant rule. Assume that a neuron's membrane time constant equals the inverse of a change in its input rate. We show that this rule is then necessary and sufficient for its subthreshold membrane potential to represent a likelihood ratio of that rate change. Thresholding that likelihood ratio makes the neuron an online detector of input rate increases and, by extension, stimulus onset when rapid reaction is required. We tested the time constant rule by reanalyzing published recordings from eight second-order sensory neurons spanning olfaction, mechanoreception, electroreception, and audition, across both insects and vertebrates. In all cases, the measured membrane time constants were consistent with the predicted inverse relationship. These examples are intended as cross-system and cross-phyla consistency checks rather than a population-level validation study. When the time constant rule holds, it yields a direct and intuitive interpretation of neural computation. A neuron's voltage is an online likelihood ratio: its time constant sets the decay of statistical evidence, its threshold is a decision line, and each output spike is a time-stamped declaration that its input statistics have changed. The rule links membrane biophysics with the statistical computation and ecological function of a neuron. Its falsifiability invites direct experimental tests to determine whether it extends more broadly across sensory pathways and neuron classes.

The proliferation of artificial neural networks (ANNs) has led to increased energy consumption, raising concerns about their sustainability. Spiking neural networks (SNNs) operate using sparse, binary spikes to communicate information between neurons and offer a potential solution due to their limited energy requirements. Another technique for reducing a neural network's footprint is quantization, which compresses weight representations to decrease memory usage and energy consumption. In this study, we present twin network augmentation (TNA), a novel training method aimed at improving the performance of SNNs on a range of benchmark data sets while also facilitating enhanced network compression through quantization of weights. TNA involves cotraining an SNN with a twin SNN with identical network architecture, optimizing both networks to minimize their cross-entropy losses and the mean squared error between their output logits. We demonstrate that TNA significantly enhances classification performance across various data sets and can be applied for reducing SNNs to ternary weight precision for inference. Our results show that TNA outperforms traditional knowledge distillation methods and achieves state-of-the-art performance for the evaluated network architecture on benchmark data sets, including CIFAR-10, CIFAR-100, and CIFAR-10-DVS. This letter underscores the effectiveness of TNA in bridging the performance gap between SNNs and ANNs and suggests further exploration into the application of TNA on different network architectures and data sets.

Working memory requires the brain to maintain information from the recent past to guide ongoing behavior. Neurons can contribute to this capacity by slowly integrating their inputs over time, creating persistent activity that outlasts the original stimulus. However, when these slowly integrating neurons are organized hierarchically, they introduce cumulative delays that create a fundamental challenge for learning: teaching signals that indicate whether behavior was correct or incorrect arrive out-of-sync with the neural activity they are meant to instruct. Here, we demonstrate that neurons enhanced with an adaptive current can compensate for these delays by responding to external stimuli prospectively, effectively predicting future inputs to synchronize with them. First, we show that such prospective neurons enable teaching signal synchronization across a range of learning algorithms that propagate error signals through hierarchical networks. Second, we demonstrate that this successfully guides learning in slowly integrating neurons, enabling the formation and retrieval of memories over extended timescales. We support our findings with a mathematical analysis of the prospective coding mechanism and learning experiments on motor control tasks. Together, our results reveal how neural adaptation could solve a critical timing problem and enable efficient learning in dynamic environments.

Hyperdimensional computing (HDC) is gaining popularity as a lightweight alternative computing paradigm, where data are represented as long vectors (commonly of dimension 10,000) and processed using simple algebraic operations. In the two-part HDC classification pipeline, data are first encoded as a high-dimensional representation, then classified based on their similarity to a set of prototypes. Conventionally, the prototypes are trained while the encoder remains static. In this work, we propose a method of training a binary/bipolar HDC encoder, relying exclusively on native HDC integer and binary operations. Evaluations with our method show an average accuracy improvement of 2.13% over the untrained encoder on several classification data sets commonly used in HDC research, with no increase in model size or inference complexity.

A key challenge in visual object recognition is developing models that generalize from limited data while maintaining transparency in their decision making. We propose a biologically inspired model that addresses both issues by classifying images based on transformation-invariant local shape key features. Following the principles of the brain's what and where pathways, each feature is encoded by an image patch and its relative location in polar coordinates, enabling interpretable and robust comparisons between inputs and class prototypes. To mimic human concept learning, prototypes are selected using clustering, improving representativeness and generalization. Results show that our model achieves human-comparable performance, with an error rate between 1% and 2% on the MNIST data set when all training images are used as prototypes. In data-limited scenarios, where only a small number of prototypes are selected, our model consistently outperforms convolutional neural networks (CNNs). To evaluate out-of-distribution generalization, we use prototypes from MNIST and test both models on the ETL-1 data set, which differs in data distribution. Although CNN accuracy drops significantly under these conditions, our model maintains high accuracy, even with few prototypes, demonstrating strong robustness and greater capacity to generalize to unseen distributions, bringing it closer to human-like recognition capabilities.

This letter shows that generative diffusion processes converge to associative memory systems at vanishing noise levels and characterizes the stability, robustness, memorization, and generation dynamics of both model classes. Morse-Smale dynamical systems are shown to be universal approximators of associative memory models, with diffusion processes as their white-noise perturbations. The universal properties of associative memory that follow are used to characterize a generic transition from generation to memory as noise diminishes. The structural stability of Morse-Smale flows-that is, the robustness of their global critical point structure-implies the stability of both trajectories and invariant measures for diffusions in the zero-noise limit. The learning and generation landscapes of these models appear as parameterized families of gradient flows and their stochastic perturbations, and the bifurcation theory for Morse-Smale systems implies that they are generically stable except at isolated parameter values, where enumerable sets of local and global bifurcations govern transitions between stable systems. These landscapes are thus characterized by ordered bifurcation sequences that create, destroy, or alter connections between rest points and are robust under small stochastic or deterministic perturbations. The framework is agnostic to model formulation, which we verify with examples from energy-based models, denoising diffusion models, and classical and modern Hopfield networks. We additionally derive structural stability criteria for Hopfield-type networks and find that simple cases violate them. Collectively, our geometric approach provides insight into the classification, stability, and emergence of memory and generative landscapes.

The action potential constitutes the digital component of the signaling dynamics of neurons. But the biophysical nature of the full-time course of the action potential associated with changes in membrane potential is mathematically distinct from its representation as a discrete set of events that encode when action potentials are triggered in a collection of spike trains. In this letter, we develop from first principles a unified functional-analytic framework for neuronal spike trains, grounded in Schwartz distribution theory. We show how this representation provides an exact operational calculus for convolution, distributional differentiation, and distributional support, which enables closed-form analysis of spike train dynamics without discretization, rate approximation, or smoothing. We then analyze the framework in the context of a two-neuron reciprocal circuit with propagation latencies and refractoriness, deriving exact results for synaptic drive, spike timing sensitivity, and causal admissibility of inputs, quantities that are either ill-defined or require approximation in conventional treatments.