Publication type: Article in scientific journal
Type of review: Peer review (publication)
Title: Learning long-term dependencies in segmented-memory recurrent neural networks with backpropagation of error
Authors: Glüge, Stefan
Böck, Ronald
Palm, Günther
Wendemuth, Andreas
DOI: 10.1016/j.neucom.2013.11.043
Published in: Neurocomputing
Volume(Issue): 141
Page(s): 54
Pages to: 64
Issue Date: 2014
Publisher / Ed. Institution: Elsevier
ISSN: 0925-2312
Language: English
Subject (DDC): 006: Special computer methods
Abstract: In general, recurrent neural networks have difficulties in learning long-term dependencies. The segmented-memory recurrent neural network (SMRNN) architecture together with the extended real-time recurrent learning (eRTRL) algorithm was proposed to circumvent this problem. Due to its computational complexity eRTRL becomes impractical with increasing network size. Therefore, we introduce the less complex extended backpropagation through time (eBPTT) for SMRNN together with a layer-local unsupervised pre-training procedure. A comparison on the information latching problem showed that eRTRL is better able to handle the latching of information over longer periods of time, even though eBPTT guaranteed a better generalisation when training was successful. Further, pre-training significantly improved the ability to learn long-term dependencies with eBPTT. Therefore, the proposed eBPTT algorithm is suited for tasks that require big networks where eRTRL is impractical. The pre-training procedure itself is independent of the supervised learning algorithm and can improve learning in SMRNN in general.
Fulltext version: Published version
License (according to publishing contract): Licence according to publishing contract
Departement: Life Sciences and Facility Management
Organisational Unit: Institute of Computational Life Sciences (ICLS)
Appears in collections:Publikationen Life Sciences und Facility Management

Files in This Item:
There are no files associated with this item.

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.