{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/learning-longer-term-dependencies-in-rnns","title":"Learning Longer-term Dependencies in RNNs with Auxiliary Losses","arxiv_id":"1803.00144","date":"2018-03-01","proceeding":"ICML 2018","authors":["Trieu H. Trinh","Andrew M. Dai","Minh-Thang Luong","Quoc V. Le"],"abstract":"Despite recent advances in training recurrent neural networks (RNNs),\ncapturing long-term dependencies in sequences remains a fundamental challenge.\nMost approaches use backpropagation through time (BPTT), which is difficult to\nscale to very long sequences. This paper proposes a simple method that improves\nthe ability to capture long term dependencies in RNNs by adding an unsupervised\nauxiliary loss to the original objective. This auxiliary loss forces RNNs to\neither reconstruct previous events or predict next events in a sequence, making\ntruncated backpropagation feasible for long sequences and also improving full\nBPTT. We evaluate our method on a variety of settings, including pixel-by-pixel\nimage classification with sequence lengths up to 16\\,000, and a real document\nclassification benchmark. Our results highlight good performance and resource\nefficiency of this approach over competitive baselines, including other\nrecurrent models and a comparable sized Transformer. Further analyses reveal\nbeneficial effects of the auxiliary loss on optimization and regularization, as\nwell as extreme cases where there is little to no backpropagation.","url_abs":"http://arxiv.org/abs/1803.00144v3","url_pdf":"http://arxiv.org/pdf/1803.00144v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"learning-longer-term-dependencies-in-rnns","repo_url":"https://github.com/younggyoseo/rnn-auxiliary-loss","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"document-classification","task_name":"Document Classification"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/sequential-image-classification-on-sequential-1","task":"Sequential Image Classification","dataset":"Sequential CIFAR-10","model":"Transformer (self-attention) (Trinh et al., 2018)","rank_in_archive_order":13,"of":13,"metrics":{"Unpermuted Accuracy":"62.2%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.00144","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}