% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@INPROCEEDINGS{Quercia:1022039,
author = {Quercia, Alessio and Morrison, Abigail and Scharr, Hanno
and Assent, Ira},
title = {{SGD} {B}iased towards {E}arly {I}mportant {S}amples for
{E}fficient {T}raining},
reportid = {FZJ-2024-01176},
pages = {TBA},
year = {2023},
abstract = {In deep learning, using larger training datasets usually
leads to more accurate models. However, simply adding more
but redundant data may be inefficient, as some training
samples may be more informative than others. We propose to
bias SGD (Stochastic Gradient Descent) towards samples that
are found to be more important after a few training epochs,
by sampling them more often for the rest of training. In
contrast to state-of-the-art, our approach requires less
computational overhead to estimate sample importance, as
itcomputes estimates once during training using the
prediction probabilities, and does not require that training
be restarted. In the experimental evaluation, we see that
our learning technique trains faster than state-of-the-art
and can achieve higher test accuracy, especially when
datasets are not well balanced. Lastly, results suggest that
our approach has intrinsic balancing properties. Code is
available at https://github.com/AlessioQuercia/sgd biased.},
month = {Dec},
date = {2023-12-01},
organization = {International Conference on Data
Mining, Shanghai (Peoples R China), 1
Dec 2023 - 4 Dec 2023},
cin = {IAS-8 / IAS-6},
cid = {I:(DE-Juel1)IAS-8-20210421 / I:(DE-Juel1)IAS-6-20130828},
pnm = {5112 - Cross-Domain Algorithms, Tools, Methods Labs (ATMLs)
and Research Groups (POF4-511) / 5232 - Computational
Principles (POF4-523) / 5234 - Emerging NC Architectures
(POF4-523) / 510 - Engineering Digital Futures –
Supercomputing, Data Management and Information Security for
Knowledge and Action (POF4-500) / HDS LEE - Helmholtz School
for Data Science in Life, Earth and Energy (HDS LEE)
(HDS-LEE-20190612)},
pid = {G:(DE-HGF)POF4-5112 / G:(DE-HGF)POF4-5232 /
G:(DE-HGF)POF4-5234 / G:(DE-HGF)POF4-510 /
G:(DE-Juel1)HDS-LEE-20190612},
typ = {PUB:(DE-HGF)8},
doi = {10.34734/FZJ-2024-01176},
url = {https://juser.fz-juelich.de/record/1022039},
}