le 19 septembre 2019
Publié le 11 février 2021 Mis à jour le 8 février 2022

Special conference Mirko Degli Esposti

photo
photo

Mirko Degli Esposti est professeur de mathématiques, Vice-recteur de l'Université de Bologne, Italie, invité par le laboratoire AGM

Recent developments of A.I. show an increasing capability of generating artificial media and art contents, with huge impact on the creativity Industry. After a brief introduction, we concentrate on automatic text generation. 

In particular, Long Short-Term Memory (LSTM) networks have recently shown remarkable performance in several tasks dealing with natural language generation, such as image captioning or poetry composition. 

Yet, only few works have analyzed text generated by LSTMs in order to quantitatively evaluate to which extent such artificial texts resemble those generated by humans. For LSTM networks a temperature-like parameter controlling the generation process shows an optimal value---for which the produced texts are closest (in a way we will explain) to real language---consistent across all the different entropic and statistical features investigated.

Date : jeudi 19 septembre 2019 de 12h30 à 14h00

Lieu : Maison internationale de la recherche, Neuville-sur-Oise.