A unified neurocomputational model for spatial and linguistic representations DOI Creative Commons
Tatsuya Haga, Yohei Oseki, Tomoki Fukai

и другие.

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown

Опубликована: Май 11, 2023

Abstract Hippocampus and entorhinal cortex encode spaces by spatially local hexagonal grid activity patterns (place cells cells), respectively. In addition, the same brain regions also implicate neural representations for non-spatial, semantic concepts (concept cells). These observations suggest that neurocomputational mechanisms spatial knowledge are related in brain. However, exact relationship remains to be understood. Here we show a mathematical correspondence between value function goal-directed navigation an information measure word embedding models natural language processing. Based on this relationship, integrate computations into representation model called as “disentangled successor information” (DSI). DSI generates biologically plausible representations: like place cells, concept-specific which resemble concept cells. Furthermore, with representations, can perform inferences of contexts words common computational framework based simple arithmetic operations. This computation interpreted partial modulations cell assemblies non-grid Our offers theoretical connection suggests possible roles hippocampal representations.

Язык: Английский

Large language models surpass human experts in predicting neuroscience results DOI Creative Commons
Xiaoliang Luo, Akilles Rechardt, Guangzhi Sun

и другие.

Nature Human Behaviour, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 27, 2024

Abstract Scientific discoveries often hinge on synthesizing decades of research, a task that potentially outstrips human information processing capacities. Large language models (LLMs) offer solution. LLMs trained the vast scientific literature could integrate noisy yet interrelated findings to forecast novel results better than experts. Here, evaluate this possibility, we created BrainBench, forward-looking benchmark for predicting neuroscience results. We find surpass experts in experimental outcomes. BrainGPT, an LLM tuned literature, performed yet. Like experts, when indicated high confidence their predictions, responses were more likely be correct, which presages future where assist humans making discoveries. Our approach is not specific and transferable other knowledge-intensive endeavours.

Язык: Английский

Процитировано

11

On the origin of memory neurons in the human hippocampus DOI Creative Commons
Luca D. Kolibius, Sheena A. Josselyn, Simon Hanslmayr

и другие.

Trends in Cognitive Sciences, Год журнала: 2025, Номер unknown

Опубликована: Март 1, 2025

Язык: Английский

Процитировано

0

Embedding Emotion Concepts in Cognitive Maps DOI Creative Commons
Yumeng Ma,

Nilofar Vafaie,

Philip A. Kragel

и другие.

Neuroscience & Biobehavioral Reviews, Год журнала: 2025, Номер unknown, С. 106089 - 106089

Опубликована: Март 1, 2025

Emotion knowledge is organized in a two-dimensional space known as the affective circumplex, which thought to develop from core feelings and co-occurrence of emotional events. Neural studies reveal that emotion concepts cognitive maps abstract are represented hippocampal-prefrontal systems. We propose circumplex formed by learning transitions between concepts, process mediated reciprocal network involving hippocampal cells encode grid medial entorhinal ventral prefrontal cortices relations them. anticipate testing this hypothesis will shed light on debate about whether emotions biologically basic or constructed dimensions.

Язык: Английский

Процитировано

0

A unified neural representation model for spatial and conceptual computations DOI Creative Commons
Tatsuya Haga, Yohei Oseki, Tomoki Fukai

и другие.

Proceedings of the National Academy of Sciences, Год журнала: 2025, Номер 122(11)

Опубликована: Март 10, 2025

The hippocampus and entorhinal cortex encode spaces by spatially local hexagonal grid activity patterns (place cells cells), respectively. In addition, the same brain regions also implicate neural representations for nonspatial, semantic concepts (concept cells). These observations suggest that neurocomputational mechanisms spatial knowledge are related in brain. However, exact relationship remains to be understood. Here, we show a mathematical correspondence between value function goal-directed navigation an information measure word embedding models natural language processing. Based on this relationship, integrate computations into representation model called “disentangled successor information” (DSI). DSI generates biologically plausible representations: like place cells, concept-specific which resemble concept cells. Furthermore, with representations, can perform inferences of contexts words common computational framework based simple arithmetic operations. This computation interpreted partial modulations cell assemblies nongrid Our offers theoretical connection suggests possible roles hippocampal representations.

Язык: Английский

Процитировано

0

Natural Concepts and the Economics of Cognition and Communication DOI Creative Commons
Peter Gärdenfors

Philosophia, Год журнала: 2024, Номер 52(4), С. 865 - 882

Опубликована: Май 8, 2024

Abstract This article takes a cognitive approach to natural concepts. The aim is introduce criteria that are evaluated with respect how they support the economy of humans when using concepts in reasoning and communicating them. I first present theory conceptual spaces as tool for expressing criteria. Then central idea correspond convex regions space. argue this criterion has far-reaching consequences regards Partly following earlier work, some other further delimit class One these coherence, which does not seem have been discussed previously. Finally, show convexity make it possible ensure people mean same thing communicate Apart from its philosophical interest, analysis presented will be relevant tasks engineering artificial systems work

Язык: Английский

Процитировано

3

The inevitability and superfluousness of cell types in spatial cognition DOI Creative Commons
Xiaoliang Luo, Robert M. Mok, Bradley C. Love

и другие.

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Янв. 12, 2024

Summary Discoveries of functional cell types, exemplified by the cataloging spatial cells in hippocampal formation, are heralded as scientific breakthroughs. We question whether identification types based on human intuitions has merit and suggest that “spatial cells” may arise non-spatial computations sufficient complexity. show deep neural networks (DNNs) for object recognition, which lack grounding, contain numerous units resembling place, border, head-direction cells. Strikingly, even untrained DNNs with randomized weights contained such support decoding information. Moreover, when these “spatial” excluded, information can be decoded from remaining DNN units, highlights superfluousness to cognition. Now large-scale simulations feasible, complexity brain should respected intuitive notions type, misleading any complex network, relegated history.

Язык: Английский

Процитировано

2

Hippocampal connectivity patterns echo macroscale cortical evolution in the primate brain DOI Creative Commons
Nicole Eichert, Jordan DeKraker, Amy Howard

и другие.

Nature Communications, Год журнала: 2024, Номер 15(1)

Опубликована: Июль 16, 2024

Abstract While the hippocampus is key for human cognitive abilities, it also a phylogenetically old cortex and paradoxically considered evolutionarily preserved. Here, we introduce comparative framework to quantify preservation reconfiguration of hippocampal organisation in primate evolution, by analysing as an unfolded cortical surface that geometrically matched across species. Our findings revealed overall conservation macro- micro-structure, which shows anterior-posterior and, perpendicularly, subfield-related organisational axes both humans macaques. However, while functional species followed axis, observed marked latter species, mirrors rudimentary integration default-mode-network non-human primates. Here show microstructurally preserved regions like may still undergo due their embedding within heteromodal association networks.

Язык: Английский

Процитировано

2

Differential reorganization of episodic and semantic memory systems in epilepsy-related mesiotemporal pathology DOI Creative Commons
Donna Gift Cabalo, Jordan DeKraker, Jessica Royer

и другие.

Brain, Год журнала: 2024, Номер 147(11), С. 3918 - 3932

Опубликована: Июль 25, 2024

Abstract Declarative memory encompasses episodic and semantic divisions. Episodic captures singular events with specific spatiotemporal relationships, whereas houses context-independent knowledge. Behavioural functional neuroimaging studies have revealed common distinct neural substrates of both systems, implicating mesiotemporal lobe (MTL) regions such as the hippocampus distributed neocortices. Here, we explored declarative system reorganization in patients unilateral temporal epilepsy (TLE) a human disease model to test impact variable degrees MTL pathology on function. Our cohort included 31 TLE 60 age- sex-matched healthy controls, all participants underwent retrieval tasks during multimodal MRI session. The were closely matched terms stimuli trial design. Capitalizing non-linear connectome gradient-mapping techniques, derived task-based topographies states, neocortical networks. Comparing hippocampal gradients between observed marked topographic systems states. Neocortical alterations characterized by reduced differentiation across lateral midline parietal cortices hemispheres. In MTL, contrast, presented more posterior anterior segments ipsilateral seizure focus pathological core, indicating perturbed intrahippocampal connectivity. Semantic was also found bilateral angular regions, unaffected. Furthermore, leveraging proxies pathology, microstructure morphology that associated TLE-related memory. Moreover, correlation analysis statistical mediation models these contributed behavioural deficits memory, but again not patients. Altogether, our findings suggest processes rely networks, are supported network involving neocortex. Alterations networks can provide compact signature state-dependent conditions damage, TLE.

Язык: Английский

Процитировано

2

Differential Reorganization of Episodic and Semantic Memory Systems in Epilepsy-Related Mesiotemporal Pathology DOI Creative Commons
Donna Gift Cabalo, Jordan DeKraker, Jessica Royer

и другие.

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown

Опубликована: Сен. 29, 2023

A bstract Declarative memory encompasses episodic and semantic divisions. Episodic captures singular events with specific spatiotemporal relationships, while houses context-independent knowledge. Behavioral functional neuroimaging studies have revealed common distinct neural substrates of both systems, implicating mesiotemporal lobe (MTL) regions distributed neocortices. Here, we studied a population patients unilateral temporal epilepsy (TLE) variable degrees MTL pathology as human disease model to explore declarative system reorganization, examine neurocognitive associations. Our cohort included 20 TLE well 60 age sex-matched healthy controls, who underwent retrieval tasks during MRI session. Tasks were closely matched in terms stimuli trial design. Capitalizing on connectome gradient mapping techniques, derived task-based topographies states, the neocortical networks. Comparing hippocampal gradients between observed topographic reorganization but not characterized by marked compression lateral midline parietal cortices hemispheres, cooccurring an expansion anterior-posterior differentiation ipsilateral pathology. These findings suggest that processes are supported network, hippocampus neocortex, such alterations can provide compact signature state-dependent conditions associated damage TLE. Leveraging microstructural morphological proxies pathology, furthermore pathological markers selective TLE-related reorganization. Moreover, correlation analysis statistical mediation models these contributed behavioral deficits patients. Altogether, our point systems TLE, which consistently affects deficits. consistent structure-function relationships reaffirm critical role systems.

Язык: Английский

Процитировано

5

The inevitability and superfluousness of cell types in spatial cognition DOI Open Access
Xiaoliang Luo, Robert M. Mok, Bradley C. Love

и другие.

Опубликована: Авг. 28, 2024

Discoveries of functional cell types, exemplified by the cataloging spatial cells in hippocampal formation, are heralded as scientific breakthroughs. We question whether identification types based on human intuitions has merit and suggest that “spatial cells” may arise non-spatial computations sufficient complexity. show deep neural networks (DNNs) for object recognition, which lack grounding, contain numerous units resembling place, border, head-direction cells. Strikingly, even untrained DNNs with randomized weights contained such support decoding information. Moreover, when these “spatial” excluded, information can be decoded from remaining DNN units, highlights superfluousness to cognition. Now large-scale simulations feasible, complexity brain should respected intuitive notions type, misleading any complex network, relegated history.

Язык: Английский

Процитировано

1