Neural Networks

Convolutional neural networks (CNNs) have become essential models for predicting neural activity and behavior in visual tasks. However, their ability to capture higher-level cognitive functions, such as numerosity discrimination, remains debated. Numerosity, the ability to perceive and estimate the number of items in a visual scene, is often proposed to rely on specialized number-detector units within CNNs, analogous to number-selective neurons observed in the brain. In this study, we use CORnet, a biologically inspired CNN architecture inspired by the organization of the primate visual system. To address a limitation of classical Representational Similarity Analysis (RSA)--its assumption that all units contribute equally--we apply pruning, a feature selection approach that identifies the units most relevant for explaining behavioral similarity structure. Our results show that number-detector units are not critical for population-level representations of numerosity, challenging their proposed role in previous studies.

Human object recognition is robust to challenging conditions, such as when ones view of an object is fragmented due to an occluding foreground object. In comparison, deep neural networks (DNNs) are typically more susceptible to occlusion, suggesting that human vision relies on distinct mechanisms. Here, we investigated the role of visual diet in the emergence of these mechanisms by asking whether human-like robustness might arise in DNNs when trained with image datasets that better reflect the properties of occlusion in natural vision. We trained convolutional and transformer DNNs to classify clear images only, images augmented with artificial occluders (i.e., geometric shapes) or natural occluders (objects segmented from photographs). We then evaluated DNN occlusion robustness and compared their performance profiles with 30 human participants. We found that DNNs trained with artificial occluders remained vulnerable to natural occlusion and exhibited less human-like performance than those trained with natural occlusion. Our findings suggest that human robustness to visual occlusion arises from learning to disentangle natural objects from each other rather than simply learning to recognize objects from partial views. They also imply that commonly used forms of artificial occlusion are unsuitable for the evaluation or promotion of robustness to real-world occlusion in DNNs.

AO_SCPLOWBSTRACTC_SCPLOWOne of the most compelling ideas for bridging neuroscience and artificial neural networks is the establishment of a framework based on three main components: network architecture, optimization mechanism, and loss (or objective) function to be minimized. While the first two components have been extensively explored, the definition of a loss or objective function in neuroscience has been addressed less thoroughly, often from perspectives such as predictive coding. In this work, we propose an elementary loss function grounded in the comparison of neuronal responses to two signals: an external one, used for learning, and an internal one, reflecting the acquired knowledge. The loss function is thus simply the basic difference between the two, which, in terms of logical signals, corresponds to a well-known non-linearly separable function: the XOR function. We illustrate with a computational example how a binarized image recognition algorithm can be straightforwardly implemented in an autoencoder, and we show how a neuronal motif organized around an inhibitory neuron could implement such XOR operation and provide a feedback signal that makes optimization possible.

Equilibrium propagation (EP) is a biologically plausible alternative to backpropagation that has demonstrated competitive performance across a range of machine learning tasks. Recent work has extended EP to spiking neural networks (SNNs), leveraging leaky integrate-and-fire (LIF) neurons and spike-based plasticity rules to improve biological realism while maintaining strong performance. In this work, we propose an EP-based SNN framework that combines LIF neural dynamics with a predictive learning rule, replacing conventional spike-timing-dependent plasticity (STDP) with a learning rule more directly aligned with predictive coding principles. We evaluate the proposed model on multiple image classification benchmarks, including MNIST, KMNIST, and Fashion-MNIST, and compare its performance with a BP-trained LIF SNN baseline. Our results show that the proposed EP-based LIF model (EP+LIF) achieves competitive accuracy across datasets, with performance approaching that of the BP-trained counterpart (BP+LIF) while relying on a biologically motivated local learning rule. In addition, analysis of hidden-layer spiking activity reveals that EP+LIF produces more persistent hidden-state activity, whereas BP+LIF yields sparser spiking representations. These results demonstrate that predictive learning can support effective EP-based training in LIF spiking networks, while also highlighting differences in activity patterns that motivate future work on activity regulation and sparse spiking dynamics.

Hippocampus is known to replay activity patterns to recall and process memories, which is often related to Hopfield-type attractor dynamics. Another line of theoretical studies suggests that hippocampal replay prioritizes replay of experiences to accelerate value learning for efficient decision making. It is unknown how hippocampal attractor dynamics perform prioritized memory sampling, and more broadly, how we can consistently relate dynamical (bottom-up) and functional (top-down) theories of hippocampal replay. In this paper, we propose an extended Hopfield-type attractor network model with momentum, kinetic energy, and conservation of the total energy, which is called momentum Hopfield model. We show that our model can be interpreted as CA3-CA1 network model with intrinsic oscillation, and such network model reproduces hippocampal replay in 1-D and 2-D spatial structures. We also prove that our model functionally works as Markov-chain Monte Carlo sampling in which recall frequencies of memory patterns can be arbitrarily biased. Using this property, we implemented prioritized experience replay using our model, which actually accelerated reinforcement learning for spatial navigation. Our model explains how dynamics of hippocampal circuits realize efficient memory sampling, providing a theoretical link between dynamics and functions of hippocampal replay.

The primary objective of system neuroscience is to understand the functional mapping and its causation in the dynamics of the brain network. Some experimental and methodological studies suggest that functional modularity and its hierarchical information processing in the brain network are crucial to understanding the functional role of task-specific or state-specific information flow in the brain. However, because most of the established techniques for detecting effective network structures in the neuroscience research field are strongly based on the "Granger causality" perspective, existing causal discovery methods specified for brain network analysis cannot identify the causal hierarchy in the modular network in the brain due to spurious correlation issues and indistinguishability of causal direction under the Gaussianity of observational noise in a linear system. To address the issues, we developed a causal discovery method for synchronous neural dynamics, called the Jacobian-informed linear non-Gaussian acyclic model, "j-VAR-LiNGAM", by incorporating the information of the Jacobian matrix determined from a phase-coupled oscillator model estimated from observed neural data into the VAR-LiNGAM algorithms. The method was validated by showing that it could extract causal ordering in both synthetic data and empirical neural observed data. Moreover, by analyzing the observed neural oscillatory signals obtained from mice and humans, we confirmed that our method identified causally hierarchical structures in the brain, which aligned with the neurophysiological interpretations. These findings suggested that our proposed method can reveal the neural basis of hierarchical information processing in the brain network.

The human visual system, and in particular the Visual Word Form Area (VWFA), adapts to process letters and words, even when the stimuli do not share canonical script features, like Braille. Here we set-up to compare the organization of typical orthographic and peculiar visual scripts such as Braille in computational models. In a first experiment, we looked at how Braille letters are represented in an illiterate Convolutional Neural Network (AlexNet) and compared them to Latin alphabet and to Line Braille, a custom line-based script. We observed a predisposition of the network, pre-trained to perform object recognition, for line-based scripts. This finding suggests an initial advantage of line junctions over Braille in processing scripts likely based on typical visual computations applied to the visual world. In a second experiment, we trained two benchmark neural network architectures (AlexNet, CORnet Z) to classify words in the Latin script (literacy acquisition) and then in the Braille script (expertise acquisition). We modelled the processing of reading visual Braille and explored the networks representations at different layers. We observed clustering of features based on the visual properties of the scripts and not by the networks expertise. Unlike human participants, the representations of linguistic categories do not converge to a model of the linguistic (orthographic, phonological, semantic) properties. Overall, the lack of alignment between the visual processing of the trained computational models and neural data recorded in expert humans suggests that the fundamental processing of reading cannot be fully explained by simple feed-forward visual processing of the script, but likely relies on additional mechanisms including interactive relations between the visual and linguistic systems.

Synaptic plasticity is metabolically expensive, yet animals continuously update their internal models without exhausting energy reserves. However, when artificial neural networks are trained, the network parameters are typically updated on every sample that is presented, even if the sample was classified correctly. Inspired by the human negativity bias and error-related negativity, we propose memorized mistake-gated learning--a biologically plausible plasticity rule where synaptic updates are strictly gated by current and past classification errors. This reduces the number of updates the network needs to make by 50% [~] 80%. Mistake gating is particularly well suited in two cases: 1) For incremental learning where new knowledge is acquired on a background of pre-existing knowledge, 2) For online learning scenarios when data needs to be stored for later replay, as mistake-gating reduces storage buffer requirements. The algorithm can be implemented in a few lines of code, adds no hyper-parameters, and comes at negligible computational overhead. Learning on mistakes is an energy efficient and biologically relevant modification to commonly used learning rules that is well suited for continual learning.

Generalization is a fundamental criterion for evaluating learning effectiveness, a domain where biological intelligence excels yet artificial intelligence continues to face challenges. In biological learning and memory, the well-documented spacing effect shows that appropriately spaced intervals between learning trials can significantly improve behavioral performance. While multiple theories have been proposed to explain its underlying mechanisms, one compelling hypothesis is that spaced training promotes integration of input and innate variations, thereby enhancing generalization to novel but related scenarios. Here we examine this hypothesis by introducing a bio-inspired spacing effect into artificial neural networks, integrating input and innate variations across spaced intervals at the neuronal, synaptic, and network levels. These spaced ensemble strategies yield significant performance gains across various benchmark datasets and network architectures. Biological experiments on Drosophila further validate the complementary effect of appropriate variations and spaced intervals in improving generalization, which together reveal a convergent computational principle shared by biological learning and machine learning.

The brain is believed to process information efficiently in a different manner from deep learning-based artificial intelligence (AI). Brain-like next-generation AI is gaining attention owing to its potential to perform human-like, highly adaptive, robust, and power-efficient computation. To realize such AI, one crucial approach is the bottom-up implementation of the neuronal systems, capturing their electrophysiological characteristics in electronic circuits. However, this neuromorphic approach generally focuses on simplified neuronal models that do not refer to many biological findings. Developing closer-to-brain models is a natural direction that serve as a fundamental computing model for next-generation AI. One of the constraints of neuromorphic circuits is the bit resolution of synaptic efficacy memory, as the memory footprint scales with it precision. Although low-resolution synaptic efficacy is essential for minimizing memory circuit footprint and energy consumption, it generally leads to performance degradation in many tasks such as the spatio-temporal spike pattern detection. This study proposed a closer-to-brain learning rule that incorporates heterosynaptic plasticity (HP) induced by glutamate spillover. It is demonstrated that our model mitigates the performance degradation associated with low-bit resolution synaptic efficacy, achieving the pattern detection success rate with 3-bit resolution synaptic efficacy, which is comparable to 64-bit floating-point precision. Furthermore, the findings of the study indicate that HP based model accelerates the convergence of the synaptic effcacy and effectively potentiates the synapses relevant to the pattern detection while suppressing irrelevant ones, thereby promoting a bimodal distribution of synaptic efficacies. These findings may provide a basic framework for constructing an energy-efficient, brain-like next-generation AI that maintains high performance under hardware constraints.

Improving long-term memory in artificial neural networks remains an open challenge. To address this, we developed a novel brain-inspired framework for memory prioritization based on the principle of emotional valence. Our framework includes: (i) a valence-weighted cross-entropy loss that scales the learning signal by the valence magnitude, analogous to neuromodulation; (ii) an amygdala-inspired module that learns high-dimensional valence embeddings; and (iii) a hippocampus-inspired module that integrates valence embeddings into the attention mechanism to modulate information retrieval. We demonstrated the generalization of our framework across spatial, episodic, and language-based memory tasks, consistently improving memory prioritization and long-term retention of high-salience information. In addition to improving long-term memory, we also showed that our framework can help mitigate the "lost-in-the-middle" problem in language modeling. More generally, this research provides further evidence of the potential of brain-inspired algorithms to advance the field of machine learning.

Recurrent neural networks (RNNs) have proven to be highly successful in emulating human-like cognitive functions such as working memory. In recent years, RNNs are evolving to incorporate more biophysical realism to produce more plausible predictions on how cognitive tasks are solved in real neural circuits. However, there are major challenges in constructing and training networks with the complex and nonlinear properties of real neurons. A major component of the success of RNNs is that they share the same mathematical base as deep neural networks, permitting highly efficient optimization of model parameters using standard deep learning techniques. To do so, they use abstract representations of neurons which fail to capture the impact of cell-level biophysical and morphologic properties that may benefit network-level function. Expanding task-trained RNNs with biophysical properties such as dendrites and active ionic currents poses substantial challenges, as it moves these models away from the validated training regimes known to be highly effective for RNNs. To address this gap, we developed a biophysically detailed reservoir computing (BRC) framework with the goal of extracting mechanistic insights from biophysical neural models, and propose that these insights can be used to guide model choices that will work for specific categories of cognitive tasks. The BRC network was constructed with synaptically coupled excitatory and inhibitory cells, in which the excitatory cells include multicompartment biophysically active dendrites; motivated by empirical studies suggesting dendrites have desirable computational benefits (e.g. pattern classification and coincidence detection). We trained the BRC network to do a simplified working memory task where it had to maintain the representation of an extrinsic "cue" input. We studied the impact of extrinsic input time constants (fast AMPA vs slow NMDA) and location (dendrite vs soma) on the ability of a network to solve the task. Our results revealed that cue inputs through NMDA receptors are particularly efficient for solving the working memory task. Further, the properties of NMDA receptors are uniquely suited for cue inputs delivered at the dendrite, as networks trained with dendritic AMPA cue inputs failed to solve the task. Detailed examination of the cell and network dynamics that solve the task reveals distinct local network configurations and computing principles for the different types of extrinsic input. Overall, much like the body of mechanistic insights that have underpinned the success of training RNNs, this study lays the groundwork for applying the BRC framework to train biophysically detailed neural models to solve complex human-like cognitive tasks.

Categorisation is often treated as a form of compression: a high-dimensional stimulus space is reduced to a smaller set of behaviourally or cognitively useful classes. However, compression alone does not determine whether a category map is useful. The present manuscript develops an information-theoretic framework for evaluating categorisation in terms of both category complexity and target-relevant information preservation. Across a set of synthetic demonstrations, alternative category maps over the same stimulus space are shown to preserve different target variables, including identity, action, nuisance, and hierarchical category structure. The framework is then extended to learned visual representations by analysing layer-derived category maps from a pretrained ResNet-50 network applied to CIFAR-10 images. Two scenarios are compared: a clean-only object run and a pooled nuisance run containing clean, blurred, pixelated, and noise-perturbed images. The results show that category maps can have substantial entropy while preserving information about a variable that is not aligned with the specified target, and that the value of a categorisation depends on the target variable to be preserved. The manuscript argues that categorisation should therefore be evaluated not only by compression or separability, but by the information retained about a specified cognitive, behavioural, or computational target.

In this study, resource-constrained learning methods were developed as a model for the learning behavior of the fly brain, specifically the mushroom body. Recent research on the mushroom bodies of flies shows that unfamiliar odors activate certain output neurons (MBONs); however, these effects are rapidly suppressed upon repeated exposure to the same odor. Such MBON behaviors appear to reflect odor learning. We investigated how flies continue learning about odors throughout their lives despite their small brains. Researchers have suggested that learning about new odors can help flies forget existing memories. Therefore, we hypothesized that the main reason for continual learning is that it serves as a strategy for forgetting. To test the validity of this hypothesis, we designed three models using a kernel perceptron. This approach is suitable for estimating ongoing learning capacity within a budget. According to the results of computer simulations and theoretical analysis, the model demonstrated the importance of forgetting mechanisms for two reasons: first, to prepare for subsequent learning sessions, and second, to reduce the negative effects of deleting memories. Author summaryDrosophila mushroom body output neurons (MBONs) in the 3 compartment of the fruit fly brain are highly activated by novel odors, and their activation triggers alerting behavior. Interestingly, these specific neurons react only to unfamiliar odor information, suggesting they constantly undergo incremental learning of new odors. This study was aimed at constructing three incremental learning models of the MBON 3 neurons. Although there have been numerous studies on complex circuit designs to reproduce activation waveforms, herein we constructed a fundamental learning model based on a kernelized learning method. Since kernelized learning models interpret Hebbian learning as the addition or subtraction of kernel functions, the model is easy to analyze theoretically. Consequently, we conclude that the forgetting property observed in the MBON 3 neurons is essential for reducing error when learning occurs within a brain of limited capacity.

Animals and humans use multiple behavioral strategies to perform tasks. However, neural implementations of multiple strategies remain elusive, as some studies propose distinct pathways, while others observe overlapping brain regions associated with strategies. We propose a hybrid deep reinforcement learning (H-DRL) method, in which one network model implements model-free and inference-based behaviors through synaptic plasticity and recurrent activity. H-DRL uses a single updating rule and switches the strategy according to task demands without an explicit arbitrator. H-DRL reproduced mixed strategies of humans in a two-step task. In the mouse perceptual decision-making task, H-DRL adapted the recurrent dynamics with rich learning when the task condition required inference-based behavior, while adopting model-free behavior with lazy learning for a simple condition. The activity of H-DRL units showed condition-dependent maintenance of previous events, consistent with orbitofrontal cortical activity in mice. Our model provides a unified view that one cortical network automatically determines strategies in use depending on task conditions.

Electrophysiological data, which serve as a biological signal that bridges neural activity and behavioral tasks, provide an innovative approach to neuroscience research. In this study, we constructed a dataset that contains over 2000 neurons across 117 days recorded in 20 mice containing 28,573 trials. Data for 5 mice were collected from the Secondary Motor Cortex (M2) region 8 mice was derived from the Ventrolateral Striatum (VLS) and 7 mice were from Substantia Nigra pars Reticulata (SNR). We induced licking behavior in head-fixed mice by periodically delivering water through a spout while simultaneously recording spiking activity from three brain regions and behavior related electrical signals. This dataset ensures precise temporal alignment between neural activity and behavioral events, offering a robust foundation for investigating neural encoding mechanisms and simulation of neural activities. This dataset establishes a precise spike-to-event mapping, which enables high decoding accuracy using Multilayer Perceptron (MLP) and Support Vector Machine (SVM). It can serve as a high-quality benchmark for developing encoding and decoding algorithms in neural networks, particularly Spiking Neural Networks (SNNs).

Understanding neural dynamics is crucial for uncovering how the brain processes information and controls behavior. Linear dynamical systems (LDS) are widely used for modeling neural data due to their simplicity and effectiveness in capturing latent dynamics. However, LDS assumes a stable mapping from the latent states to neural activity, limiting its ability to capture representational drift--gradual changes in the brains representation of the external world. To address this, we introduce the Stiefel Manifold Dynamical System (SMDS), a new class of model designed to account for drift in neural representations across trials. In SMDS, emission matrices are constrained to be orthonormal and evolve smoothly over trials on the Stiefel manifold--the space of all orthonormal matrices--while the dynamics parameters are shared. This formulation allows SMDS to leverage data across trials while accounting for non-stationarity, thus capturing the underlying neural dynamics more accurately compared to an LDS. We apply SMDS to both simulated datasets and neural recordings across species. Our results consistently show that SMDS outperforms LDS in terms of log-likelihood and requires fewer latent dimensions to capture the same activity. Moreover, SMDS provides a powerful framework for quantifying and interpreting representational drift. It reveals a gradual drift over the course of minutes in the neural recordings and uncovers varying drift rates across dimensions, with slower drift in behaviorally and neurally significant dimensions.

Convolutional neural networks (CNNs) are a class of artificial neural networks (ANNs). Since the early 2010s, they have been widely adopted as models of primate vision and classifiers of neuroimaging data, becoming relevant for a wealth of neuroscientific fields. However, the majority of neuroscience researchers come from soft-science backgrounds (like medicine, biology, or psychology) and do not have enough quantitative skills to understand the inner workings of A/CNNs. To avoid undesirable black boxes, neuroscientists should acquire some rudiments of computational neuroscience and machine learning. However, most researchers do not have the time nor the resources to make big learning investments, and self-study materials are hardly tailored to people with little mathematical background. This paper aims to fill this gap by providing a concise but accurate introduction to CNNs and their use in neuroscience -- using the minimum required mathematics, neuroscientific analogies, and Python code examples. A companion Jupyter Notebook guides readers through code examples, translating theory into practice and providing visual outputs. The paper is organised in three sections: The Concepts, The Implementation, and The Biological Plausibility of A/CNNs. The three sections are largely independent, so readers can either go through the entire paper or select a section of interest.

A central goal of neuroscience is to understand how neural circuit architecture supports learning. While recent work has clarified the computational role of depth in sensory cortical hierarchies, it remains unclear why predominantly feedforward, non-convolutional circuits such as the cerebellum and olfactory system also contain multiple processing layers. Theoretical work in deep learning has shown that two-hidden-layer networks can achieve classification capacity that scales quadratically with the number of intermediate neurons, but these results rely on nonlocal synaptic optimization and are therefore difficult to reconcile with biological learning rules. Here, we show analytically and numerically that a two-hidden-layer network with feedforward gating can achieve quadratic capacity using local three-factor Hebbian learning when intermediate activity is sparse. This architecture supports efficient one-shot learning and, in settings where backpropagation requires many repeated weight updates, offers an advantage in learning speed. Beyond random perceptron tasks, the model also performs well on structured cerebellum-related tasks, including reinforcement-learning-based motor control. Mapping the model onto cerebellar microcircuitry further suggests functional roles for dendritic compartmentalization, branch-specific inhibition, and disinhibitory interneuron pathways. Together, these results extend the Marr-Albus-Ito framework by showing how the presence of multiple intermediate layers in cerebellum-like circuits can support fast, local, and high-capacity learning.

As social beings, humans make decisions partly based on social interaction. Observing the behavior of others can lead to learning from and about them, potentially increasing trust and prompting trust-based behavioral changes. Observation-based decision making involves different neural structures. The orbitofrontal cortex (OFC) and lateral prefrontal cortex (LPFC) are known as neural structures mainly involved in processing emotional and cognitive decision values, respectively, while the anterior cingulate cortex (ACC) plays a pivotal role as a social hub, integrating the afferent expectancy signals from OFC and LPFC. This paper presents a neurocomputational model of the interplay between observational learning and trust, as well as their role in individual decision-making. Our model elucidates and predicts the emotional and rational behavioral changes of an individual influenced by observing the action-outcome association of an alleged expert. We have modeled the neurodynamics of three cortical structures (OFC, LPFC, and ACC) and their interactions, where the neural oscillatory properties, modeled with Dynamic Bayesian Probability, represent the observers attitude towards the expert and the decision options. As an example of an everyday behavioral situation related to climate change, we use the choice of transportation between home and work. The EEG-like simulation outputs from our model represent the presumed brain activity of an individual making such a choice, assuming the decision-maker is exposed to social information.