Among neurons of the same type, the different electrophysiological parameters vary drastically. It is presently unclear if there is any hidden structure in this diversity and how it relates to neuronal function. One potential source of structure is the evolutionary pressure that has sculpted brains to work in an energy-efficient manner. Here we show that neurons' parameters do not vary randomly, but that neurons populate a manifold within their parameter space permitting energy-efficient signaling. This manifold represents a Pareto front of degenerate solutions to the problem of energy-efficiency and explains the prevalence of low to intermediate firing rates (2 - 5 Hz) in neocortex. Furthermore, we show that neurons in different sensory brain areas populate different manifolds and that food restriction induces systematic shifts of neuronal parameters along this manifold. Our results suggest that the large diversity of neuronal properties is actually tightly regulated to ensure energy-efficient signaling in different contexts.

Abstract Active sensing systems are known to adapt the structure of sensory signals. Whether they can improve perception by controlling when sensory information is acquired remains unclear. We show that echolocating fruit bats exposed to rhythmic noise preferentially emit calls during recurring low-noise periods, a behaviour we term “dip echolocation”. Dip echolocation occurred in laboratory and wild bats and represents an active-sensing analogue of dip listening in humans. A normative model showed that temporal positioning of calls emerges from a trade-off between sensory information and energetic cost, alongside concurrent adaptations of call structure. Pharmacological inactivation of the frontal auditory field disrupted precise temporal control, implicating a role for frontal cortical circuits in adaptive vocal timing. These findings identify adaptive vocal timing as an active-sensing strategy for overcoming acoustic interference.

The efficient coding hypothesis states that biological perceptual systems adapt to the statistics of the sensory signals arising in their natural environments. Because infants actively shape these sensory statistics through their own behavior, perception and action form a tightly coupled developmental loop. We present an integrative review of recent extensions of efficient coding into the domain of active perception, with a particular focus on the Active Efficient Coding (AEC) framework. AEC explains the development of perceptions and actions through a unifying computational principle: encoding sensory observations as efficiently as possible. We introduce a novel formalism that frames AEC within rate-distortion theory, interpreting active perception as a problem of lossy compression. We then re-examine AEC models of the autonomous learning and calibration of active binocular vision in a simulated infant embodiment. This work shows how active perception can emerge without external supervision, providing a foundation for the development of complex behaviors and higher cognition.

Motion retargeting from humans to human-like artificial agents is becoming increasingly important as humanoid robots grow more capable. However, most existing approaches focus only on reproducing kinematics and ignore the rich sensorimotor experience associated with human movement. In this work, we present a framework for simulating the multimodal sensorimotor experiences of infants using physical and virtual humanoids. From a single video, our method reconstructs the infant's body configuration by extracting its skeletal structure and estimating the full 3D pose from each frame. Then we map the reconstructed motion onto several developmental platforms: the physical iCub robot and the virtual simulators pyCub, EMFANT and MIMo. Replaying the retargeted motions on these embodiments produces simulated multisensory streams including proprioception (joints and muscles), touch, and vision. For the best-matching embodiment, the retargeting achieves sub-centimeter accuracy and enables a rich multimodal analysis of infant development as well as enhanced automated annotation of behaviors. This framework provides a unique window into the infant's sensorimotor experience, offering new tools for robotics, developmental science, and early detection of neurodevelopmental disorders. The code is available at https://github.com/ctu-vras/motion-retargeting/.

Rolling over is one of the earliest milestones in infant motor development, reflecting the emergence of coordinated, whole-body sensorimotor control. Here, we conduct a computational study of infant rolling using MIMo, a virtual infant embodiment equipped with proprioception and vestibular sensation. MIMo learns supine-to-prone rolls with reinforcement learning. Interestingly, the learned behaviors capture developmental trends and coordination patterns consistent with those reported in real infants, including improved performance and faster execution with age. Our results explain how infant capabilities and constraints can give rise to realistic behaviors in artificial agents, with a particular emphasis on how motor development is shaped by the changing body morphology. This work highlights the role of embodied computational models as a powerful tool for studying sensorimotor development.

Facial attractiveness has been linked to statistical regularities such as symmetry and averageness, suggesting that beauty may depend on the ease with which a face is perceived. We empirically test this hypothesis by training variational autoencoders on four face datasets without attractiveness supervision and evaluating their representations on the 597 faces from the Chicago Face Database. Across models, human attractiveness ratings closely aligns with the direction defined by the VAE evidence lower bound (ELBO) in rate-distortion space. Independently learned latent spaces contain an attractiveness direction that transfers strongly across random initializations and training data. We also find that attractive faces are more prototypical in both shape and latent space. Our results connect classic accounts of aesthetics with learned generative models and provide empirical support for a variational interpretation of the processing fluency theory of aesthetic pleasure.

The ability to predict the future is of great value for biological and artificial cognitive systems alike. However, successfully predicting the future typically requires maintaining a memory of the recent past. It is currently unclear how biological or artificial spiking neural networks can learn to maintain past sensory information to help predict the future. Here we propose Predictive Coding Light+ (PCL+), a spiking neural network architecture for unsupervised sequence processing that learns recurrent excitatory connections with delays to enable short-term retention of information. We show that the PCL+ network reproduces classic findings on sequence learning in visual cortex. Furthermore, it learns to “fill in” missing input in a challenging gesture recognition task. Overall, our work shows how spiking neural networks can learn recurrent excitatory connections with delays to maintain a record of the recent past and successfully predict the future.

Humans acquire semantic object representations from egocentric visual streams with minimal supervision. Importantly, the visual system processes with high resolution only the center of its field of view and learns similar representations for visual inputs occurring close in time. This emphasizes slowly changing information around gaze locations. This study investigates the role of central vision and slowness learning in the formation of semantic object representations in humans. We simulate five months of human-like visual experience using the Ego4D dataset and generate gaze coordinates with a state-of-the-art gaze prediction model. Using these predictions, we extract crops that mimic central vision and train a time-contrastive Self-Supervised Learning model on them. Our results show that combining temporal slowness and central vision improves the encoding of different semantic facets of object representations. Specifically, focusing on central vision strengthens the extraction of foreground object features, while considering temporal slowness, especially during fixational eye movements, allows the model to encode broader semantic information about objects. These findings provide new insights into the mechanisms by which humans may develop semantic object representations from natural visual experience. Our code will be made public upon acceptance.

In the mobile conjugate reinforcement paradigm, an infant's leg is connected to a mobile via a string, allowing the infant to move the mobile via moving their leg. Over a few minutes, infants exhibit an increase in the frequency of movement of the connected leg. This behavior is sometimes interpreted as an indication that infants experience the efficacy of causal control. However, some researchers have argued that an underlying causal model is not necessary and that a simple reinforcement model that favors mobile movements can explain this behavioral pattern. Interestingly, after the mobile is disconnected from the leg, some infants transiently show an even higher frequency of movement, a phenomenon known as the extinction burst, that is hard to reconcile with a simple reinforcement learning model alone. In this study, we propose different computational models and study to what extent they are capable of capturing infants' behavior. In particular, we construct an active-learning causal model that is capable of discovering the underlying cause-effect relationship on the fly without the need to specify either the cause or the effect in advance. We also propose an active-learning mechanism based on expectation violation, that can be combined with the proposed causal model and a number of alternative models, including a naïve reinforcement model, to give rise to an extinction burst. Overall, our work sheds light on possible learning mechanisms giving rise to infant's developing understanding of cause and effect relationships. SUMMARY: A causality-driven model is proposed that successfully simulates infant behavior in the mobile paradigm. The causal model actively discovers the underlying causal mechanism, while alternative models fail under specific simulated conditions. An active-learning mechanism based on expectation violation is introduced that unifies causal learning and hypothesis-testing within a single, coherent framework. We show that this mechanism successfully models the extinction burst and demonstrates robustness across different experimental settings.

OBJECTIVE:Despite decades of research, the circuit mechanisms that underlie focal epileptogenesis remain incompletely understood. In this study, we aimed to characterize the changes in hippocampal network activity induced by an epileptogenic insult. METHODS:In rats, prolonged electrical perforant pathway stimulation leads in a high percentage of animals to the development of temporal lobe epilepsy. However, a subset of animals remained apparently non-epileptic for unknown reasons. We monitored alterations of neuronal activity by chronically recording the local field potential (LFP) from the hippocampal dentate gyrus before, during, and after the potentially epileptogenic insult and quantifying the time-dependent changes in the periodic and the aperiodic components of the LFP. RESULTS:It is intriguing that animals with epilepsy identified by subsequent spontaneous epileptic seizures were characterized by a transient increase in the aperiodic exponent, suggesting a shift toward a reduced local excitation-to-inhibition (E/I) ratio during the first days after the perforant path stimulation. Furthermore, these animals developed a strong impairment of theta oscillation prevalence and regularity during early epileptogenesis. In contrast, resilient non-epilepsy animals without spontaneous seizures showed neither this modulation in E/I ratio nor a corruption of hippocampal theta activity. In fact, the increase in the aperiodic exponent on the first day after completion of the electrical stimulation paradigm could predict epileptogenesis with very high fidelity (area under the curve [AUC] 0.92) and correlated significantly with later seizure rate. SIGNIFICANCE:The very early phase of epileptogenesis seems to be dominated by a relative dominance of inhibitory synaptic activity in affected circuits. This finding opens the opportunity to dissect mechanisms of epileptogenesis and to test the effectiveness of anti-epileptogenesis treatment in very early disease stages by allowing identification of individuals at high risk. Furthermore, it might offer a potential explanation for the frequently observed failure of anti-epileptogenesis drugs boosting γ-aminobutyric acid (GABA)ergic inhibition.

There is strong and diverse evidence for mental rotation (MR) abilities in adults. However, current evidence for MR in children rests on just a few behavioral paradigms adapted from the adult literature. Here, we leverage recent computational models of the development of children's object recognition abilities to re-assess the evidence for MR in children. The computational models simulate infants' acquisition of object representations during embodied interactions with objects. We consider two different object recognition strategies, different from MRs, and assess their ability to replicate results from three classical MR tasks assigned to children between the ages of 6 months and 5 years. Our results show that MR may play no role in producing the results obtained from children younger than 5 years. In fact, we find that a simple recognition strategy that reflects a pixel-wise comparison of stimuli is sufficient to model children's behavior in the most used MR task. Thus, our study reopens the debate on how and when children develop genuine MR abilities. SUMMARY: We use a bio-inspired machine learning models to examine evidence of mental rotation abilities in children We show that a simple recognition strategy suffices to solve habituation-based tasks used to assess mental rotations in young children. We demonstrate that a model forming expectations about a rotation transformation can explain children's results in mental rotation tasks with the violation-of-expectation paradigm. The investigated recognition strategies no longer work when the mental rotation tasks are made closer to adults'.

Toddlers learn to recognize objects from different viewpoints with almost no supervision. During this learning, they execute frequent eye and head movements that shape their visual experience. It is presently unclear if and how these behaviors contribute to toddlers' emerging object recognition abilities. To answer this question, we here combine head-mounted eye tracking during dyadic play with unsupervised machine learning. We approximate toddlers' central visual field experience by cropping image regions from a head-mounted camera centered on the current gaze location estimated via eye tracking. This visual stream feeds a neural network model, which uses a biologically plausible unsupervised learning objective. Our experiments demonstrate that a few minutes of such first-person experience suffice to learn strong object representations permitting invariant object recognition. Importantly, by simulating alternative gaze behaviors we show that toddlers' eye movement patterns play a crucial role in this. Our analysis also reveals that the limited size of the central visual field where visual acuity is high plays an important role for successful learning. Together, this highlights the benefits of temporally structured visual experience arising from toddlers' natural interactions with objects. SUMMARY: We combine recordings of toddlers' first-person central visual field experience with biologically inspired self-supervised learning algorithms to model toddlers' development of invariant object recognition. Just a few minutes of toddlers' central visual field experience captured with head-mounted eye tracking suffice to learn strong object representations. Simulated alternative gaze behaviors produce weaker representations, demonstrating the importance of toddlers' active gaze strategies for learning. Our results emphasize the importance of toddlers' eye movements for learning object representations.

While human infants robustly discover their own causal efficacy, standard reinforcement learning agents remain brittle, as their reliance on correlation-based rewards fails in noisy, ecologically valid scenarios. To address this, we introduce the Causal Action Influence Score (CAIS), a novel intrinsic reward rooted in causal inference. CAIS quantifies an action's influence by measuring the 1-Wasserstein distance between the learned distribution of sensory outcomes conditional on that action, p(h|a), and the baseline outcome distribution, p(h). This divergence provides a robust reward that isolates the agent's causal impact from confounding environmental noise. We test our approach in a simulated infant-mobile environment where correlation-based perceptual rewards fail completely when the mobile is subjected to external forces. In stark contrast, CAIS enables the agent to filter this noise, identify its influence, and learn the correct policy. Furthermore, the high-quality predictive model learned for CAIS allows our agent, when augmented with a surprise signal, to successfully reproduce the "extinction burst" phenomenon. We conclude that explicitly inferring causality is a crucial mechanism for developing a robust sense of agency, offering a psychologically plausible framework for more adaptive autonomous systems.

We present Hierarchical Residual Networks (HiResNets), deep convolutional neural networks with long-range residual connections between layers at different hierarchical levels. HiResNets draw inspiration on the organization of the mammalian brain by replicating the direct connections from subcortical areas to the entire cortical hierarchy. We show that the inclusion of hierarchical residuals in several architectures, including ResNets, results in a boost in accuracy and faster learning. A detailed analysis of our models reveals that they perform hierarchical compositionality by learning feature maps relative to the compressed representations provided by the skip connections.

Current machine learning systems consume vastly more energy than biological brains. Neuromorphic systems aim to overcome this difference by mimicking the brain's information coding via discrete voltage spikes. However, it remains unclear how both artificial and natural networks of spiking neurons can learn energy-efficient information processing strategies. Here we propose Predictive Coding Light (PCL), a recurrent hierarchical spiking neural network for unsupervised representation learning. In contrast to previous predictive coding approaches, PCL does not transmit prediction errors to higher processing stages. Instead it suppresses the most predictable spikes and transmits a compressed representation of the input. Using only biologically plausible spike-timing based learning rules, PCL reproduces a wealth of findings on information processing in visual cortex and permits strong performance in downstream classification tasks. Overall, PCL offers a new approach to predictive coding and its implementation in natural and artificial spiking neural networks.

Cortical networks are capable of unsupervised learning and spontaneous replay of complex temporal sequences. Endowing artificial spiking neural networks with similar learning abilities remains a challenge. In particular, it is unresolved how different plasticity rules can contribute to both learning and the maintenance of network stability during learning. Here we introduce a biologically inspired form of GABA-Modulated Spike Timing-Dependent Plasticity (GMS) and demonstrate its ability to permit stable learning of complex temporal sequences including natural language in recurrent spiking neural networks. Motivated by biological findings, GMS utilizes the momentary level of inhibition onto excitatory cells to adjust both the magnitude and sign of Spike Timing-Dependent Plasticity (STDP) of connections between excitatory cells. In particular, high levels of inhibition in the network cause depression of excitatory-to-excitatory connections. We demonstrate the effectiveness of this mechanism during several sequence learning experiments with character- and token-based text inputs as well as visual input sequences. We show that GMS maintains stability during learning and spontaneous replay and permits the network to form a clustered hierarchical representation of its input sequences. Overall, we provide a biologically inspired model of unsupervised learning of complex sequences in recurrent spiking neural networks.

This paper introduces Canalizing Babbling, a development-inspired approach for data collection in sensorimotor learning. The method draws inspiration from reflexes in newborns, which are here hypothesized to scaffold the acquisition of coordinated sensorimotor actions by facilitating the early experience of contingent sensory and motor events. In the presented approach, a visual saliency system selects targets in 3D space, guiding an inverse dynamics controller to generate coordinated movements across multiple body parts of the MIMo simulated agent. Statistical analysis shows that the visual and motor observations collected using Canalizing Babbling exhibit a higher degree of coordination compared to those obtained through non-curated strategies. These findings suggest that biologically inspired exploration techniques like Canalizing Babbling can lower sample complexity, potentially accelerate downstream learning in embodied agents, and provide a framework to further investigate the role of reflexes in developmental learning.

In this article, we review a class of neuro-mimetic computational models that we place under the label of spiking predictive coding. Specifically, we review the general framework of predictive processing in the context of neurons that emit discrete action potentials, i.e., spikes. Theoretically, we structure our survey around how prediction errors are represented. This results in an organization into three broad classes of approaches: prediction errors in explicit groups of error neurons, in membrane potentials, and implicit prediction error encoding. Furthermore, we examine some applications of spiking predictive coding that utilize more energy-efficient, edge-computing hardware platforms. Finally, we highlight important future directions and challenges in this emerging line of inquiry in brain-inspired computing. Building on the prior results of work in computational cognitive neuroscience, machine intelligence, and neuromorphic engineering, we hope that this review of neuromorphic formulations and implementations of predictive coding will encourage and guide future research and development in this emerging research area.

Adapting movements to rapidly changing conditions is fundamental for interacting with our dynamic environment. This adaptability relies on internal models that predict and evaluate sensory outcomes to adjust motor commands. Even infants anticipate object properties for efficient grasping, suggesting the use of internal models. However, how internal models are adapted in early childhood remains largely unexplored. This study investigated a naturalistic force adaptation task in 1.5-, 3-year-olds, and young adults. Participants opened a drawer with temporarily increased resistance, creating sensory prediction errors between predicted and actual drawer dynamics. After perturbation, all age groups showed lower peak speed, longer movement time, and more movement units with trial-wise changes analyzed as adaptation process. Results revealed no age differences in adapting peak speed and movement units, but 1.5- and 3-year-olds exhibited higher trial-to-trial variability and were slower in adapting their movement time, although they also adapted their movement time more strongly. Upon removal of perturbation, we found significant aftereffects across all age groups, indicating effective internal model adaptation. These results suggest that even 1.5-year-olds form internal models of force parameters and adapt them to reduce sensory prediction errors, possibly through more exploration and with more variable movement dynamics compared to adults.