March 1, 1994

使用梯度下降学习长期依赖关系是困难的

Key Points

Key points are not available for this paper at this time.

Abstract

循环神经网络可以用于将输入序列映射到输出序列，例如用于识别、生成或预测问题。然而，已经报告了在训练循环神经网络执行输入/输出序列中存在长时间间隔的时序相关任务时的实际困难。我们展示了随着需捕捉依赖关系持续时间的增加，基于梯度的学习算法面临的难题日益加剧。这些结果揭示了梯度下降高效学习与长期保存信息之间的权衡。基于对该问题的理解，考虑了标准梯度下降的替代方案。

Connected Papers

Building similarity graph...

Analyzing shared references across papers

Discussion

Authors

Yoshua Bengio

P. Simard

Paolo Frasconi

Journals

IEEE Transactions on Neural Networks

Actions

Institutions

Université de Montréal

AT&T (United States)

University of Florence

References and Citations

Connected Papers

Building similarity graph...

Analyzing shared references across papers

Discussion

Cite this study

Bengio 等人（星期二，）研究了这一问题。

www.synapsesocial.com/papers/69027ac529206953ba14953e — DOI: https://doi.org/10.1109/72.279181

Also consider

Synapse has enriched 4 closely related papers on similar clinical questions. Consider them for comparative context:

Improving the convergence of back-propagation learning with second-order methods· 1989 · 333 citations
Optimization by Simulated Annealing· 1983 · 44,438 citations
Advances in Neural Information Processing Systems 5· 1993 · 1,857 citations
Parallel Distributed Processing· 1986 · 15,330 citations

使用梯度下降学习长期依赖关系是困难的

Key Points

Abstract

Citation Network

Connected Papers

Discussion

Authors

Journals

Actions

Institutions

References and Citations

Citation Network

Connected Papers

Discussion

Cite this study

Also consider