Multi-Domain Learning (MDL) refers to the problem of learning a set of models derived from a common deep architecture, each one specialized to perform a task in a certain domain (e.g., photos, sketches, paintings). This paper tackles MDL with a particular interest in obtaining domain-specific models with an adjustable budget in terms of the number of network parameters and computational complexity. Our intuition is that, as in real applications the number of domains and tasks can be very large, an effective MDL approach should not only focus on accuracy but also on having as few parameters as possible. To implement this idea we derive specialized deep models for each domain by adapting a pre-trained architecture but, differently from other methods, we propose a novel strategy to automatically adjust the computational complexity of the network. To this aim, we introduce Budget-Aware Adapters that select the most relevant feature channels to better handle data from a novel domain. Some constraints on the number of active switches are imposed in order to obtain a network respecting the desired complexity budget. Experimentally, we show that our approach leads to recognition accuracy competitive with state-of-the-art approaches but with much lighter networks both in terms of storage and computation.

Budget-Aware Adapters for Multi-Domain Learning / Berriel, Rodrigo; Stephane, Lathuiliere; Nabi, Moin; Klein, Tassilo; Oliveira-Santos, Thiago; Sebe, Niculae; Ricci, Elisa. - (2019), pp. -382. (Intervento presentato al convegno IEEE Comference on Computer Vision (ICCV'19) tenutosi a Seoul nel October 27-November 2, 2019).

Budget-Aware Adapters for Multi-Domain Learning

Stephane Lathuiliere;Moin Nabi;Nicu Sebe;Elisa Ricci
2019-01-01

Abstract

Multi-Domain Learning (MDL) refers to the problem of learning a set of models derived from a common deep architecture, each one specialized to perform a task in a certain domain (e.g., photos, sketches, paintings). This paper tackles MDL with a particular interest in obtaining domain-specific models with an adjustable budget in terms of the number of network parameters and computational complexity. Our intuition is that, as in real applications the number of domains and tasks can be very large, an effective MDL approach should not only focus on accuracy but also on having as few parameters as possible. To implement this idea we derive specialized deep models for each domain by adapting a pre-trained architecture but, differently from other methods, we propose a novel strategy to automatically adjust the computational complexity of the network. To this aim, we introduce Budget-Aware Adapters that select the most relevant feature channels to better handle data from a novel domain. Some constraints on the number of active switches are imposed in order to obtain a network respecting the desired complexity budget. Experimentally, we show that our approach leads to recognition accuracy competitive with state-of-the-art approaches but with much lighter networks both in terms of storage and computation.
2019
IEEE Conference on Computer Vision
New York
IEEE
978-1-7281-4803-8
Berriel, Rodrigo; Stephane, Lathuiliere; Nabi, Moin; Klein, Tassilo; Oliveira-Santos, Thiago; Sebe, Niculae; Ricci, Elisa
Budget-Aware Adapters for Multi-Domain Learning / Berriel, Rodrigo; Stephane, Lathuiliere; Nabi, Moin; Klein, Tassilo; Oliveira-Santos, Thiago; Sebe, Niculae; Ricci, Elisa. - (2019), pp. -382. (Intervento presentato al convegno IEEE Comference on Computer Vision (ICCV'19) tenutosi a Seoul nel October 27-November 2, 2019).
File in questo prodotto:
File Dimensione Formato  
Berriel_Budget-Aware_Adapters_for_Multi-Domain_Learning_ICCV_2019_paper.pdf

accesso aperto

Tipologia: Post-print referato (Refereed author’s manuscript)
Licenza: Altra licenza (Other type of license)
Dimensione 844.45 kB
Formato Adobe PDF
844.45 kB Adobe PDF Visualizza/Apri
09009060.pdf

Solo gestori archivio

Tipologia: Versione editoriale (Publisher’s layout)
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 806.83 kB
Formato Adobe PDF
806.83 kB Adobe PDF   Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11572/250828
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 21
  • ???jsp.display-item.citation.isi??? 17
social impact