Why is Deep Learning so cool?

Giornata dell'Associazione Italiana per la ricerca in Computer Vision, Pattern Recognition e Machine Learning, GIRPR → VPL

Modena, 29 gennaio 2018
Università di Modena e Reggio Emilia, Dipartimento di Ingegneria "Enzo Ferrari"
Sala Eventi del Tecnopolo di Modena




Programma

  • 10:00 Opening, poster presentations

    Prof. Alessandro Capra, Direttore Dipartimento di Ingegneria "Enzo Ferrari"
    Prof. Rita Cucchiara, AImageLab UNIMORE

  • 10:30 Talk by Prof. Naftali Tishby, Hebrew University of Jerusalem: "Information Theory of Deep Learning"*
    Chair: Prof. Costantino Grana, UNIMORE

    Abstract:I will present a novel comprehensive theory of large scale learning with Deep Neural Networks, based on the correspondence between Deep Learning and the Information Bottelneck framework. The new theory has the following components: (1) rethinking Learning theory; I will prove a new generalization bound, the input-compression bound, which shows that compression of the representation of input variable is far more important for good generalization than the dimension of the network hypothesis class, an ill defined notion for deep learning. (2) I will prove that for large scale Deep Neural Networks the mutual information on the input and the output variables, for the last hidden layer, provide a complete characterization of the sample complexity and accuracy of the network. This makes the information Bottelneck bound for the problem as the optimal trade-off between sample complexity and accuracy with ANY learning algorithm. (3) I will show how Stochastic Gradient Descent, as used in Deep Learning, achieves this optimal bound. In that sense, Deep Learning is a method for solving the Information Bottlneck problem for large scale supervised learning problems. The theory provide a new computational understating of the benefit of the hidden layers, and gives concrete predictions for the structure of the layers of Deep Neural Networks and their design principles. These turn out to depend solely on the joint distribution of the input and output and on the sample size.
    *This talk is partially supported by the Doctorate School in ICT of UNIMORE ( the phd students which need an attendance declaration should send an email to lorenzo.baraldi@unimore.it)

  • 11:30 Panel: the GIRPR → VPL speaker corner

    Deep Learning, Pattern Recognition and Vision, which direction?
    With contributions from the members of the association
    Chair: Dr. Simone Calderara, UNIMORE

    Speakers:

    • Prof. Nicu Sebe, University of Trento
    • Dr. Lamberto Ballan, University of Padova
    • Prof. Emanuele Frontoni, Università  Politecnica delle Marche
    • Prof. Davide Maltoni, University of Bologna
    • Prof. Marco Cristani, University of Verona
    • Prof. Barbara Caputo, Sapienza Rome University
  • 13:00 Pranzo e presentazione dei poster

  • 14:00 Assemblea straordinaria del GIRPR

  • 16:30 Presentazione dei poster e discussione finale


Iscrizione

Per iscriversi all'evento, è richiesta la compilazione del form di registrazione al seguente link.

Sono disponibili anche le informazioni per l'iscrizione all'associazione al seguente link.


Come arrivare

L'evento si terrà presso la Sala Eventi del Tecnopolo di Modena, in via P. Vivarelli, 10 a Modena. E' disponibile una mappa da cui ottenere le indicazioni stradali.


Informazioni

Per informazioni contattare il Dott. Lorenzo Baraldi (local chair) e la Dott.ssa Silvia Caliò (secretary) via e-mail o allo 059-2058774