Skip to main content
Weiter zur Homepage der Europäischen Kommission (öffnet in neuem Fenster)
Deutsch de
CORDIS - Forschungsergebnisse der EU
CORDIS

A computational neuroscience encoding-decoding approach for explaining and comparing artificial and biological networks

Projektbeschreibung

Computational Neuroscience zur Förderung der erklärbaren KI

Die jüngsten Fortschritte im Bereich der künstlichen Intelligenz stützen sich stark auf tiefe künstliche neuronale Netze und maschinelles Lernen. Diese Technologien finden breite Anwendung in den Bereichen Verkehr, Energieverteilung sowie medizinische Diagnose. Das Gebiet der erklärbaren KI (XAI) entstand mit dem Ziel, KI-Entscheidungen transparenter zu machen. Bislang hat es noch Schwierigkeiten, dies effektiv zu tun. Das über die Marie-Skłodowska-Curie-Maßnahme unterstützte Projekt Neurosci-ANN wird neuartige Techniken der Computational Neuroscience anwenden, um XAI voranzubringen und entscheidende Erkenntnisse zu gewinnen, die Forschenden helfen könnten, KI und Robotik zu revolutionieren.

Ziel

Recent progress in artificial intelligence (AI) has been mostly due to machine learning and, in particular, deep artificial neural networks (ANNs). Deep learning has an increasing presence in everyday life, including critical applications such as medical diagnosis, transportation, and energy distribution. In response to this, the field of Explainable AI (XAI) has generated much effort in terms of techniques and algorithms to address this problem. However, there is still no consensus on a suite of technology to address these challenges, progress has been extremely limited, and the formal properties of such systems are under-studied.

On the other hand, computational neuroscience (CNS) aims to discover the principles behind biological neural networks that enable the brain to support cognition, perception, and action. This project will employ the latest approaches and techniques used in the field of CNS to develop the field of XAI. Specifically, the first major goal will be to employ the methods of representational geometry and neural encoding manifolds (both proven to be effective in revealing meaningful neural relationships in previous studies) to reveal how activations of collections of artificial neurons in hidden layers are associated with the decision-making process of deep networks.

Second, the same methodology will be used to reveal novel insights from a variety of existing large-scale biological datasets. Finally, we will compare and contrast the encoding strategies of neural populations found various deep learning architectures with those observed in biological networks. A better understanding of the inner-workings of biological models could directly inform researchers on how to build novel artificial models that are more accurate, robust, and even economical during both training and inference in terms of data, time, and energy consumption.

Wissenschaftliches Gebiet (EuroSciVoc)

Schlüsselbegriffe

Programm/Programme

Thema/Themen

Finanzierungsplan

HORIZON-TMA-MSCA-PF-EF - HORIZON TMA MSCA Postdoctoral Fellowships - European Fellowships

Alle im Rahmen dieses Finanzierungsinstruments finanzierten Projekte anzeigen

Aufforderung zur Vorschlagseinreichung

(öffnet in neuem Fenster) HORIZON-MSCA-2024-PF-01

Alle im Rahmen dieser Aufforderung zur Einreichung von Vorschlägen finanzierten Projekte anzeigen

Koordinator

IE UNIVERSIDAD
Netto-EU-Beitrag
€ 209 914,56
Adresse
CALLE CARDENAL ZUNIGA 12
40003 Segovia
Spanien
Region
Centro (ES) Castilla y León Segovia
Aktivitätstyp
Higher or Secondary Education Establishments
Links
Gesamtkosten
Keine Daten