Skip to main content
Aller à la page d’accueil de la Commission européenne (s’ouvre dans une nouvelle fenêtre)
français fr
CORDIS - Résultats de la recherche de l’UE
CORDIS

Generative understanding of memory and planning

Objectif

Deep generative networks offer a step change in understanding the neural mechanisms supporting human cognition. These networks, trained to predict their own inputs, learn abstract conceptual representations, and schema for translating them into imagined experience, forming a predictive model of the world.

GENERATIVE-MP will build from the “complementary learning systems” idea, of hippocampus storing episodic experience and replaying it to consolidate semantic knowledge in neocortex, by treating neocortex as a deep generative network. In this view, memories can be efficiently stored in hippocampus as a conceptual outline together with episode-unique details, because the concepts are learned to enable neocortical (re)construction of episodes. We will test three fundamental predictions via psychological and functional neuroimaging experiments while people experience virtual reality, stories and artificial stimuli:
1). Memories will become more schematic, with fewer veridical details, as retrieval moves from hippocampus to neocortex with consolidation;
2) Efficient hippocampal episodic representations should group together conceptually similar memories, and change less to encode episodes containing predictable details;
3) Consolidation of sequential experience captures predictive relationships between the elements of experience to enable planning of movements between them. If these relationships have a 2D (map-like) structure, grid cell-like representations in neocortex will support such planning.
GENERATIVE-MP will use the results to develop a hippocampal-generative neocortical model capable of simulating and interpreting them.

The outcome will be a comprehensive understanding of episodic memory and consolidation of semantic and conceptual knowledge for planning, implemented in a neural network that processes realistic stimuli. This will provide a platform for understanding how specific sets of stimuli are processed, e.g. to facilitate learning and avoid pitfalls.

Champ scientifique (EuroSciVoc)

Ce projet n'a pas encore été classé par EuroSciVoc.
Proposez les domaines scientifiques qui vous semblent les plus pertinents et aidez-nous à améliorer notre service de classification.

Mots‑clés

Programme(s)

Thème(s)

Régime de financement

HORIZON-ERC - HORIZON ERC Grants

Voir tous les projets financés dans le cadre de ce programme de financement

Appel à propositions

(s’ouvre dans une nouvelle fenêtre) ERC-2025-ADG

Voir tous les projets financés au titre de cet appel

Institution d’accueil

UNIVERSITY COLLEGE LONDON
Contribution nette de l'UE
€ 2 200 828,42
Adresse
GOWER STREET
WC1E 6BT London
Royaume-Uni
Région
London Inner London — West Camden and City of London
Type d’activité
Higher or Secondary Education Establishments
Liens
Coût total
Aucune donnée

Bénéficiaires (1)