Neural representations

Similarity between neural representations is often quantified by measuring alignment of the representations over a set of natural stimuli that are relatively far apart in stimulus space (Kriegeskorte et al. 2008). However, systems with similar global structure can have strikingly different sensitivities to local stimulus distortions (Szegedy et al. 2013), suggesting a need for metrics that compare local sensitivities of representations. We propose a framework for comparing a set of image representations in terms of their sensitivities to local distortions. We quantify the local geometry of a r

…provide valuable models for studying neural representations of complex sounds. In this study, we recorded zebra finches' neural activity from the auditory pallium during song processing using a movable electrode array, collecting data from 51 recording sites. We employed stacked bidirectional long short-term memory (BiLSTM) deep neural networks to decode the amplitude envelope and time-locked envelope features of the songs. To evaluate h

Primate visual object recognition is supported by critical neural circuits housed in the inferior temporal (IT) cortex. How similar is one macaque IT to another? This study investigates inter-individual variation in IT cortical representations by analyzing neural responses to 640 visual stimuli (100 ms presentation at 8 degrees) across eight macaque monkeys. We employed two complementary analytical approaches: a direct one-to-one neuron matching method to assess exact neural correspondence and a many-to-one linear regression framework to evaluate potential linear transformations between indivi

Memories are thought to be stored in synapses and retrieved through the reactivation of neural ensembles. However, neural representations exhibit significant dynamics even in stable environments, a phenomenon called representational drift. Theoretical studies suggest that multiple neural representations can correspond to a memory, with post-learning exploration of these representation solutions driving drift. However, the functional implications of drift-explored representations versus those initially learned remain poorly understood. Here we show that drift uncovers robust, sparsely engaged r

Understanding how neural representations align between biological and artificial systems has emerged as a central challenge in computational neuroscience. While deep neural networks now reliably predict neural responses across multiple brain areas, their utility for understanding biological computation remains limited by our ability to accurately measure representational similarities. This limitation becomes particularly acute when working with sparse neural recordings, where traditional similarity metrics may fail to capture true representational relationships. Recent spectral analyses of sim

Animals rely on sensory inputs to navigate complex, dynamic environments. Understanding how the brain integrates these inputs to guide behavior remains a central question in neuroscience. Training AI and animals on the same task offers a unique opportunity to uncover both similarities and critical differences in the strategies used by biological and artificial systems. By pinpointing what the brain accomplishes that modern AI lacks, we can drive advancements in machine learning, pushing AI to more robust, generalizable solutions. In this study, we train mice and reinforcement learning agents o

…similarity between high-dimensional neural representations is an active research in computational neuroscience. Such similarity measures have implications for understanding variability in neural computations across individuals and species [1], as well as for comparisons between biological systems and computational models [2]. One common approach to understand these systems is to build regression models or “decoders” that reconstruct features of the stimulus from neural responses. Here, we leverage this idea to quantify the similarity of

Even in stable environments, sensory responses undergo continuous reformatting, a phenomenon known as representational drift. Using chronic calcium imaging in mouse auditory cortex, we show that during this representational drift signal correlations predict future noise correlations, suggesting that stimulus-driven co-activation strengthens effective connectivity via Hebbian-like plasticity. Linear network models reveal that these temporal dependencies between signal and noise correlations emerge only when Hebbian learning balances stochastic synaptic changes, preventing functional degradation

Over the past decade, neural representations have been studied from the lens of low-dimensional subspaces defined by the co-activation of neurons. However, this view has overlooked other forms of covarying structure in neural activity, including i) condition-specific high-dimensional neural sequences, and ii) representations that change over time due to learning or drift. In this talk, I will present a new framework that extends the classic view towards additional types of covariability that are not constrained to a fixed, low-dimensional subspace. In addition, I will present sliceTCA, a new t

Over the past decade, neural representations have been studied from the lens of low-dimensional subspaces defined by the co-activation of neurons. However, this view has overlooked other forms of covarying structure in neural activity, including i) condition-specific high-dimensional neural sequences, and ii) representations that change over time due to learning or drift. In this talk, I will present a new framework that extends the classic view towards additional types of covariability that are not constrained to a fixed, low-dimensional subspace. In addition, I will present sliceTCA, a new t

Previous work investigating the neural dynamics underlying context-dependent decision making typically analyses a single brain region or recurrent neural network (RNN) [1,2]. However, evidence suggests that the information required to solve tasks is distributed across multiple regions [3]. Here, we investigate the neural dynamics across seven brain regions of the non-human primate brain where such distributed information has been observed[3]. By examining within-region geometry and dynamics, we identified significant differences not captured by classical decoding analyses. Using multi-regional

…olfaction and language related brain’s neural resources are not well understood [1]. Behavioral task: the data that we used to compare our model output were collected from a Swedish National Study on Aging and Care in Kungsholmen (SNAC-K), where ~2500 subjects underwent an odor naming test of

…highly similar inputs into distinct neural representations. This function is mainly attributed to the sparsely firing excitatory granule cells. It is not known how excitatory DG neurons modify their firing patterns in response to different conditions, to facilitate learning. We studied the activity of excitatory DG neurons during discrimination fear conditioning task (FC) using in vivo one-photon calcium imaging in adult, healthy C57/black mice. We compared learning in males (n = 6) and females (n = 6) and further, in voluntarily running

…perceived negative emotion, but the neural mechanisms are unclear. This study aims to investigate brain responses underlying the efficacy of these interventions, by combining self-reported anxiety measures and fMRI data.Prior to scanning, personalised, anxiety-provoking scenar

…technical challenge as exploring the neural underpinnings of self-initiated behaviors in a quantitative and rigorous manner is difficult without experimenter-defined trial structures. Although loneliness has been declared an epidemic, there are no therapeutic targets and no mechanistic explanations for how social isolation negatively impacts our health. The dynamic time course of social isolation has been conceptualized in a model of social homeostasis (Lee et al., 2021). Multiple timescales for

…stability and preserved integrity of neural representations. In extreme cases, such as psychotic states in schizophrenia, there is a disruption of this stability, often resulting in the fusion of distinct neural representations. In this study, both familiar (F) and novel (N) linear mazes were utilized to assess the quality of neural coding across various time scales in 3-6 months old rats (N = 12). Lower (0.1 mg/kg) and higher doses (0.15 mg/kg) of MK-801 were administered. We observed disrupted spat

…architecture is to investigate neural representations across different subregions, e.g. by immediate early gene imaging, which visualizes genomic responses of activated neuronal popula

Prior research indicates that while the fidelity of neural representations in the visuocortical system supporting selective attention remains intact in late adulthood, it substantially diminishes in individuals with mild cognitive impairment (MCI), a transitional stage preceding dementia. However, it remains unclear to what extent neural substrates underlying attentional deficits vary across various MCI subtypes, commonly classified based on Montreal Cognitive Assessment (MoCA) performance. Here, employing an inverted encoding model (IEM) approach, we reconstructed spatially selective attentio

…Neural representations in the medial entorhinal cortex (MEC) constitute a generalized map of the local space, but recent findings may suggest that the MEC possesses non-spatial dimensions suggesting a generalized network for learning state spaces irrespectiv

The dominant theoretical framework to account for reinforcement learning in the brain is temporal difference (TD) reinforcement learning. The TD framework predicts that some neuronal elements should represent the reward prediction error (RPE), which means they signal the difference between the expected future rewards and the actual rewards. The prominence of the TD theory arises from the observation that firing properties of dopaminergic neurons in the ventral tegmental area appear similar to those of RPE model-neurons in TD learning. Previous implementations of TD learning assume a fixed temp

Cortical circuits embody remarkably reliable neural representations of sensory stimuli that are critical for perception and action. The fundamental structure of these network representations is thought to arise early in development prior to the onset of sensory experience. However, how these endogenously generated networks respond to the onset of sensory experience, and the extent to which they reorganize with experience remains unclear. Here we examine this ‘nature-nurture transform’ using chronic in vivo calcium imaging to probe the developmental emergence of the representation of orientatio

Social memory, one's ability to recognize and remember experiences with other conspecifics, consists of several discrete psychological processes. These include the ability to rapidly detect whether an individual is novel or familiar (“I know that person, but from where?”) and the recollection of an individual’s specific identity and the associated set of past experiences (“ah, we met at COSYNE last year”). These processes have conflicting demands and requirements. Familiarity is an abstract concept that can readily generalize to any individual in any context. In contrast, social episodes consi

Animals identify odors based on the combinatorial activation of olfactory sensory neurons having distinct chemical receptive fields. However, as the dose response of any olfactory sensory neuron is believed to follow a sigmoidal function that grows monotonically until it reaches a plateau, the subset of olfactory sensory neurons activated by an odor is expected to scale up with the odor concentration, undercutting the efficiency of combinatorial coding. Here we show that olfactory sensory neurons can in fact undergo a silent state upon strong and prolonged excitation called depolarization bloc

Navigating rodents show “forward sweeps” - sequential reactivation of non-local neural representations of future states starting from their current position. While direct evidence remains scant, it is widely thought that such forward sweeps support model-based (MB) evaluation of actions. However, a key challenge in testing this hypothesis is the absence of experimental paradigms that allow for precise measurement of MB action evaluation. Notably, typical assessments of MB evaluation, i.e. flexibility following changes to rewards (reward revaluation), conflate reliance on full MB evaluation wit