One common loss function in neural network classification tasks is
Categorical Cross Entropy (CCE), which punishes all misclassifications equally.
However, classes often have an inherent structure. For instance, classifying an
image of a rose as "violet" is better than as "truck". We introduce SimLoss, a
drop-in replacement for CCE that incorporates class similarities along with two
techniques to construct such matrices from task-specific knowledge. We test
SimLoss on Age Estimation and Image Classification and find that it brings
significant improvements over CCE on several metrics. SimLoss therefore allows
for explicit modeling of background knowledge by simply exchanging the loss
function, while keeping the neural network architecture the same. Code and
additional resources can be found at https://github.com/konstantinkobs/SimLoss.
%0 Conference Paper
%1 kobs2020simloss
%A Kobs, Konstantin
%A Steininger, Michael
%A Zehe, Albin
%A Lautenschlager, Florian
%A Hotho, Andreas
%B Foundations of Intelligent Systems
%C Cham
%D 2020
%I Springer International Publishing
%K accepted myown
%P 431--439
%R 10.1007/978-3-030-59491-6_41
%T SimLoss: Class Similarities in Cross Entropy
%U https://doi.org/10.1007/978-3-030-59491-6_41
%X One common loss function in neural network classification tasks is
Categorical Cross Entropy (CCE), which punishes all misclassifications equally.
However, classes often have an inherent structure. For instance, classifying an
image of a rose as "violet" is better than as "truck". We introduce SimLoss, a
drop-in replacement for CCE that incorporates class similarities along with two
techniques to construct such matrices from task-specific knowledge. We test
SimLoss on Age Estimation and Image Classification and find that it brings
significant improvements over CCE on several metrics. SimLoss therefore allows
for explicit modeling of background knowledge by simply exchanging the loss
function, while keeping the neural network architecture the same. Code and
additional resources can be found at https://github.com/konstantinkobs/SimLoss.
%@ 978-3-030-59491-6
@inproceedings{kobs2020simloss,
abstract = {One common loss function in neural network classification tasks is
Categorical Cross Entropy (CCE), which punishes all misclassifications equally.
However, classes often have an inherent structure. For instance, classifying an
image of a rose as "violet" is better than as "truck". We introduce SimLoss, a
drop-in replacement for CCE that incorporates class similarities along with two
techniques to construct such matrices from task-specific knowledge. We test
SimLoss on Age Estimation and Image Classification and find that it brings
significant improvements over CCE on several metrics. SimLoss therefore allows
for explicit modeling of background knowledge by simply exchanging the loss
function, while keeping the neural network architecture the same. Code and
additional resources can be found at https://github.com/konstantinkobs/SimLoss.},
added-at = {2020-04-14T15:04:52.000+0200},
address = {Cham},
author = {Kobs, Konstantin and Steininger, Michael and Zehe, Albin and Lautenschlager, Florian and Hotho, Andreas},
biburl = {https://www.bibsonomy.org/bibtex/20303f0b5558a822fe25df4eec20d07db/msteininger},
booktitle = {Foundations of Intelligent Systems},
description = {SimLoss: Class Similarities in Cross Entropy},
doi = {10.1007/978-3-030-59491-6_41},
eventtitle = {ISMIS 2020},
interhash = {c9693bde2183871e64852bd39bcf396b},
intrahash = {0303f0b5558a822fe25df4eec20d07db},
isbn = {978-3-030-59491-6},
keywords = {accepted myown},
pages = {431--439},
publisher = {Springer International Publishing},
timestamp = {2022-10-27T08:22:14.000+0200},
title = {SimLoss: Class Similarities in Cross Entropy},
url = {https://doi.org/10.1007/978-3-030-59491-6_41},
year = 2020
}