Training Neural Networks in Single vs Double Precision

Journal
14th International Conference on Knowledge Discovery and Information Retrieval
ISSN
2184-3228
ISBN
978-989-758-614-9
Type
conference paper
Date Issued
2022-10
Author(s)
;
;
DOI
Abstract
The commitment to single-precision floating-point arithmetic is widespread in the deep learning community. To evaluate whether this commitment is justified, the influence of computing precision (single and double precision) on the optimization performance of the Conjugate Gradient (CG) method (a second-order optimization algorithm) and RMSprop (a first-order algorithm) has been investigated. Tests of neural networks with one to five fully connected hidden layers and moderate or strong nonlinearity with up to 4 million network parameters have been optimized for Mean Square Error (MSE). The training tasks have been set up so that their MSE minimum was known to be zero. Computing experiments have disclosed that single-precision can keep up (with superlinear convergence) with double-precision as long as line search finds an improvement. First-order methods such as RMSprop do not benefit from double precision. However, for moderately nonlinear tasks, CG is clearly superior. For strongly nonlinear tasks, both algorithm classes find only solutions fairly poor in terms of mean square error as related to the output variance. CG with double floating-point precision is superior whenever the solutions have the potential to be useful for the application goal.
Language
English
HSG Classification
contribution to scientific community
Refereed
Yes
Book title
Proceedings of the 14th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2022) - KDIR
Publisher
SciTePress
Start page
307
End page
314
Pages
8
Event Title
14th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2022) - KDIR
Event Location
Valetta, Malta
Event Date
24-26 October 2022
Official URL
URL
Contact Email Address
bernhard.bermeitinger@unisg.ch