Skip to main content

A Unified Framework of Surrogate Loss by Refactoring and Interpolation

Author(s): Liu, Lanlan; Wang, Mingzhe; Deng, Jia

To refer to this page use:
Full metadata record
DC FieldValueLanguage
dc.contributor.authorLiu, Lanlan-
dc.contributor.authorWang, Mingzhe-
dc.contributor.authorDeng, Jia-
dc.identifier.citationLiu, Lanlan, Mingzhe Wang, and Jia Deng. "A Unified Framework of Surrogate Loss by Refactoring and Interpolation." In European Conference on Computer Vision (2020): pp. 278-293. doi:10.1007/978-3-030-58580-8_17en_US
dc.description.abstractWe introduce UniLoss, a unified framework to generate surrogate losses for training deep networks with gradient descent, reducing the amount of manual design of task-specific surrogate losses. Our key observation is that in many cases, evaluating a model with a performance metric on a batch of examples can be refactored into four steps: from input to real-valued scores, from scores to comparisons of pairs of scores, from comparisons to binary variables, and from binary variables to the final performance metric. Using this refactoring we generate differentiable approximations for each non-differentiable step through interpolation. Using UniLoss, we can optimize for different tasks and metrics using one unified framework, achieving comparable performance compared with task-specific losses. We validate the effectiveness of UniLoss on three tasks and four datasets. Code is available at
dc.format.extent278 - 293en_US
dc.relation.ispartofEuropean Conference on Computer Visionen_US
dc.rightsAuthor's manuscripten_US
dc.titleA Unified Framework of Surrogate Loss by Refactoring and Interpolationen_US
dc.typeConference Articleen_US

Files in This Item:
File Description SizeFormat 
UnifiedFramework.pdf505.79 kBAdobe PDFView/Download

Items in OAR@Princeton are protected by copyright, with all rights reserved, unless otherwise indicated.