L’équipe du GRAL est heureuse d’annoncer la conférence du Prof. McClelland, le vendredi 26-09-2014 à la Fondation Universitaire, à partir de 18h (http://www.fondationuniversitaire.be/en/contact).

Au regard du nombre de places limité, vous êtes invités à vous inscrire auprès d’Angélique Bernacki (abernack@ulb.ac.be).

Argumentaire

« How experience leads to readiness to learn »

Research on human development and learning reveals many examples of cases in which children of one age readily learn something, while children a bit younger do not. When children are ready, we say that they are in ‘the zone of proximal development’ — but how can we explain when and how they enter this zone? My research using artificial neural network models has explored this and related issues — including striking U-shaped developmental trends, in which learning appears to go backwards before it progresses. These issues arise in a range of different contexts, including children’s readiness to learn such things as basic intuitions about principles of physics, the conceptual organization of different natural kinds, and the correct interpretation of mathematical symbols, such at the equal sign. In my talk I will mention these cases, and discuss how learning processes operating in deep neural networks can help us understand these phenomena.