Automated learning rates in machine learning for dynamic mini-batch sub-sampled losses

Show simple item record

dc.contributor.advisor Wilke, Daniel Nicolas
dc.contributor.postgraduate Kafka, Dominic
dc.date.accessioned 2024-04-18T09:17:34Z
dc.date.available 2024-04-18T09:17:34Z
dc.date.created 2021
dc.date.issued 2020-12
dc.description Thesis (PhD (Mechanical Engineering))--University of Pretoria, 2020. en_ZA
dc.description.abstract Learning rate schedule parameters remain some of the most sensitive hyperparameters in machine learning, as well as being challenging to resolve, in particular when mini-batch subsampling is considered. Mini-batch sub-sampling (MBSS) can be conducted in a number of ways, each with their own implications on the smoothness and continuity of the underlying loss function. In this study, dynamic MBSS, often applied in approximate optimization, is considered for neural network training. For dynamic MBSS, the mini-batch is updated for every function and gradient evaluation of the loss and gradient functions. The implication is that the sampling error between mini-batches changes abruptly, resulting in non-smooth and discontinuous loss functions. This study proposes an approach to automatically resolve learning rates for dynamic MBSS loss functions using gradient-only line searches (GOLS) over fifteen orders of magnitude. A systematic study is performed, which investigates the characteristics and the influence of training algorithms, neural network architectures and activation functions on the ability of GOLS to resolve learning rates. GOLS are shown to compare favourably against the state-ofthe-art probabilistic line search for dynamic MBSS loss functions. Matlab and PyTorch 1.0 implementations of GOLS are available for both practical training of neural networks as well as a research tool to investigate dynamic MBSS loss functions. en_ZA
dc.description.availability Unrestricted en_ZA
dc.description.degree PhD (Mechanical Engineering) en_ZA
dc.description.department Mechanical and Aeronautical Engineering en_ZA
dc.identifier.citation * en_ZA
dc.identifier.other S2021 en_ZA
dc.identifier.uri http://hdl.handle.net/2263/95641
dc.language.iso en en_ZA
dc.publisher University of Pretoria
dc.rights © 2021 University of Pretoria. All rights reserved. The copyright in this work vests in the University of Pretoria. No part of this work may be reproduced or transmitted in any form or by any means, without the prior written permission of the University of Pretoria.
dc.subject UCTD en_ZA
dc.subject Machine learning en_ZA
dc.title Automated learning rates in machine learning for dynamic mini-batch sub-sampled losses en_ZA
dc.type Thesis en_ZA


Files in this item

This item appears in the following Collection(s)

Show simple item record