@inproceedings{5f748f6202c148ffb10c415767119e53,
title = "Tree Memory Networks for Sequence Processing",
abstract = "Long-term dependencies are difficult to learn using Recurrent Neural Networks due to the vanishing and exploding gradient problems, since their hidden transform operation is applied linearly in sequence length. We introduce a new layer type (the Tree Memory Unit), whose weight application scales logarithmically in the sequence length. We evaluate this on two pathologically hard memory benchmarks and two datasets. On those three tasks which require long-term dependencies, it strongly outperforms Long Short-Term Memory baselines. However, it does show weaker performance on sequences with few long-term dependencies. We believe that our approach can lead to more efficient sequence learning if used on sequences with long-term dependencies.",
author = "Frederik Diehl and Alois Knoll",
note = "Publisher Copyright: {\textcopyright} 2019, Springer Nature Switzerland AG.; 28th International Conference on Artificial Neural Networks, ICANN 2019 ; Conference date: 17-09-2019 Through 19-09-2019",
year = "2019",
doi = "10.1007/978-3-030-30487-4_34",
language = "English",
isbn = "9783030304867",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "431--443",
editor = "Tetko, {Igor V.} and Pavel Karpov and Fabian Theis and Vera Kurkov{\'a}",
booktitle = "Artificial Neural Networks and Machine Learning – ICANN 2019",
}