A Deep Incremental Boltzmann Machine for Modeling Context in Robots

Fethiye Irmak Dogan, Hande Celikkanat, Sinan Kalkan

Tutkimustuotos: Artikkeli kirjassa/raportissa/konferenssijulkaisussaKonferenssiartikkeliTieteellinenvertaisarvioitu

Kuvaus

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.
Alkuperäiskielienglanti
Otsikko2018 IEEE International Conference on Robotics and Automation (ICRA)
Sivumäärä6
KustantajaIEEE
Julkaisupäivä13 syyskuuta 2018
Sivut2411-2416
ISBN (painettu)978-1-5386-3082-2
ISBN (elektroninen)978-1-5386-3081-5
DOI - pysyväislinkit
TilaJulkaistu - 13 syyskuuta 2018
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisuussa
TapahtumaIEEE International Conference on Robotics and Automation -
Kesto: 21 toukokuuta 201825 toukokuuta 2018

Tieteenalat

  • 113 Tietojenkäsittely- ja informaatiotieteet

Lainaa tätä

Dogan, F. I., Celikkanat, H., & Kalkan, S. (2018). A Deep Incremental Boltzmann Machine for Modeling Context in Robots. teoksessa 2018 IEEE International Conference on Robotics and Automation (ICRA) (Sivut 2411-2416). IEEE. https://doi.org/10.1109/ICRA.2018.8462925
Dogan, Fethiye Irmak ; Celikkanat, Hande ; Kalkan, Sinan. / A Deep Incremental Boltzmann Machine for Modeling Context in Robots. 2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2018. Sivut 2411-2416
@inproceedings{fd6d4accf05e4069af09ded45fe12684,
title = "A Deep Incremental Boltzmann Machine for Modeling Context in Robots",
abstract = "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.",
keywords = "113 Computer and information sciences",
author = "Dogan, {Fethiye Irmak} and Hande Celikkanat and Sinan Kalkan",
year = "2018",
month = "9",
day = "13",
doi = "10.1109/ICRA.2018.8462925",
language = "English",
isbn = "978-1-5386-3082-2",
pages = "2411--2416",
booktitle = "2018 IEEE International Conference on Robotics and Automation (ICRA)",
publisher = "IEEE",
address = "International",

}

Dogan, FI, Celikkanat, H & Kalkan, S 2018, A Deep Incremental Boltzmann Machine for Modeling Context in Robots. julkaisussa 2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE, Sivut 2411-2416, IEEE International Conference on Robotics and Automation, 21/05/2018. https://doi.org/10.1109/ICRA.2018.8462925

A Deep Incremental Boltzmann Machine for Modeling Context in Robots. / Dogan, Fethiye Irmak; Celikkanat, Hande; Kalkan, Sinan.

2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2018. s. 2411-2416.

Tutkimustuotos: Artikkeli kirjassa/raportissa/konferenssijulkaisussaKonferenssiartikkeliTieteellinenvertaisarvioitu

TY - GEN

T1 - A Deep Incremental Boltzmann Machine for Modeling Context in Robots

AU - Dogan, Fethiye Irmak

AU - Celikkanat, Hande

AU - Kalkan, Sinan

PY - 2018/9/13

Y1 - 2018/9/13

N2 - 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.

AB - 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.

KW - 113 Computer and information sciences

U2 - 10.1109/ICRA.2018.8462925

DO - 10.1109/ICRA.2018.8462925

M3 - Conference contribution

SN - 978-1-5386-3082-2

SP - 2411

EP - 2416

BT - 2018 IEEE International Conference on Robotics and Automation (ICRA)

PB - IEEE

ER -

Dogan FI, Celikkanat H, Kalkan S. A Deep Incremental Boltzmann Machine for Modeling Context in Robots. julkaisussa 2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE. 2018. s. 2411-2416 https://doi.org/10.1109/ICRA.2018.8462925