A Deep Incremental Boltzmann Machine for Modeling Context in Robots

Fethiye Irmak Dogan, Hande Celikkanat, Sinan Kalkan

Forskningsoutput: Kapitel i bok/rapport/konferenshandlingKonferensbidragVetenskapligPeer review

Sammanfattning

Context is an essential capability for robots that are to be as adaptive as possible in challenging environments. Although there are many context modeling efforts, they assume a fixed structure and number of contexts. In this paper, we propose an incremental deep model that extends Restricted Boltzmann Machines. Our model gets one scene at a time, and gradually extends the contextual model when necessary, either by adding a new context or a new context layer to form a hierarchy. We show on a scene classification benchmark that our method converges to a good estimate of the contexts of the scenes, and performs better or onpar on several tasks compared to other incremental models or non-incremental models.
Originalspråkengelska
Titel på gästpublikation2018 IEEE International Conference on Robotics and Automation (ICRA)
Antal sidor6
FörlagIEEE
Utgivningsdatum13 sep 2018
Sidor2411-2416
ISBN (tryckt)978-1-5386-3082-2
ISBN (elektroniskt)978-1-5386-3081-5
DOI
StatusPublicerad - 13 sep 2018
MoE-publikationstypA4 Artikel i en konferenspublikation
EvenemangIEEE International Conference on Robotics and Automation -
Varaktighet: 21 maj 201825 maj 2018

Vetenskapsgrenar

  • 113 Data- och informationsvetenskap

Citera det här