Imperial College London
Portrait Picture

Dr Ben Glocker

Faculty of Engineering, Department of Computing

Professor in Machine Learning for Imaging
 
 
 
//

Contact

 

+44 (0)20 7594 8334b.glocker Website CV

 
 
//

Location

 

377Huxley BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@inproceedings{Li:2019:10.1007/978-3-030-32248-9_45,
author = {Li, Z and Kamnitsas, K and Glocker, B},
doi = {10.1007/978-3-030-32248-9_45},
pages = {402--410},
publisher = {Springer Verlag},
title = {Overfitting of neural nets under class imbalance: Analysis and improvements for segmentation},
url = {http://dx.doi.org/10.1007/978-3-030-32248-9_45},
year = {2019}
}
Download

RIS format (EndNote, RefMan)

TY  - CPAPER
AB  - Overfitting  in  deep  learning  has  been  the  focus  of  a  num-ber  of  recent  works,  yet  its  exact  impact  on  the  behaviour  of  neuralnetworks is not well understood. This study analyzes overfitting by ex-amining  how  the  distribution  of  logits  alters  in  relation  to  how  muchthe model overfits. Specifically, we find that when training with few datasamples, the distribution of logit activations when processing unseen testsamples  of  an  under-represented  class  tends  to  shift  towards  and  evenacross the decision boundary, while the over-represented class seems un-affected.  In  image  segmentation,  foreground  samples  are  often  heavilyunder-represented.  We  observe  that  sensitivity  of  the  model  drops  asa result of overfitting, while precision remains mostly stable. Based onour analysis, we derive asymmetric modifications of existing loss func-tions  and  regularizers  including  a  large  margin  loss,  focal  loss,  adver-sarial training and mixup, which specifically aim at reducing the shiftobserved when embedding unseen samples of the under-represented class.We study the case of binary segmentation of brain tumor core and showthat  our  proposed  simple  modifications  lead  to  significantly  improvedsegmentation performance over the symmetric variants.
AU  - Li,Z
AU  - Kamnitsas,K
AU  - Glocker,B
DO  - 10.1007/978-3-030-32248-9_45
EP  - 410
PB  - Springer Verlag
PY  - 2019///
SN  - 0302-9743
SP  - 402
TI  - Overfitting of neural nets under class imbalance: Analysis and improvements for segmentation
UR  - http://dx.doi.org/10.1007/978-3-030-32248-9_45
UR  - https://link.springer.com/chapter/10.1007%2F978-3-030-32248-9_45
UR  - http://hdl.handle.net/10044/1/72113
ER  -
Download