Digits that are not: Generating new types through deep neural nets

Akın Kazakçı 1 Cherti Mehdi 2 Balázs Kégl 3, 2, 4
4 TAO - Machine Learning and Optimisation
LRI - Laboratoire de Recherche en Informatique, UP11 - Université Paris-Sud - Paris 11, Inria Saclay - Ile de France, CNRS - Centre National de la Recherche Scientifique : UMR8623
Abstract : For an artificial creative agent, an essential driver of the search for novelty is a value function which is often provided by the system designer or users. We argue that an important barrier for progress in creativity research is the inability of these systems to develop their own notion of value for novelty. We propose a notion of knowledge-driven creativity that circumvent the need for an externally imposed value function, allowing the system to explore based on what it has learned from a set of referential objects. The concept is illustrated by a specific knowledge model provided by a deep generative au-toencoder. Using the described system, we train a knowledge model on a set of digit images and we use the same model to build coherent sets of new digits that do not belong to known digit types.
Type de document :
Communication dans un congrès
International Conference on Computational Creativity, Jun 2016, Paris, France. 〈http://www.computationalcreativity.net/iccc2016/proceedings-2016/〉
Liste complète des métadonnées

Littérature citée [35 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-01427556
Contributeur : Akin Osman Kazakci <>
Soumis le : jeudi 5 janvier 2017 - 18:10:05
Dernière modification le : lundi 12 novembre 2018 - 10:57:20
Document(s) archivé(s) le : jeudi 6 avril 2017 - 14:47:26

Fichier

digits (17) (1).pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01427556, version 1

Citation

Akın Kazakçı, Cherti Mehdi, Balázs Kégl. Digits that are not: Generating new types through deep neural nets. International Conference on Computational Creativity, Jun 2016, Paris, France. 〈http://www.computationalcreativity.net/iccc2016/proceedings-2016/〉. 〈hal-01427556〉

Partager

Métriques

Consultations de la notice

619

Téléchargements de fichiers

308