LSTM model introduction

Séminaire
  • Quand ? 12/05/2016 à partir de 15:00 (America/Montreal / UTC-400)
  • Où ? Au local D4-2011 de la Faculté des sciences
  • Nom du contact
  • Participants CONFÉRENCIER : Stanislas Lauly, étudiant au doctorat en informatique
  • Ajouter l'événement au calendrier iCal

Langue de présentation : Anglais

Résumé : Modeling structured sequences have always been an important but also difficult challenge. Recently, neural network models are achieving increasingly good results in predicting time series, having the ability to model long term dependencies well. The most successful approach right now is the LSTM (Long Short-Term Memory) model, providing state-of-the-art results for a lot of different tasks. For this talk, I will present in details how the LSTM works and continue with a simpler and faster version called the GRU (Gated Recurrent Unit) model.

Date : Le jeudi 12 mai 2016

Heure : 15 h

Local : D4-2011

Toutes les personnes sont cordialement invitées.