{"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/regularizing-rnns-for-caption-generation-by","title":"Regularizing RNNs for Caption Generation by Reconstructing The Past with The Present","arxiv_id":"1803.11439","date":"2018-03-30","proceeding":"CVPR 2018 6","authors":["Xinpeng Chen","Lin Ma","Wenhao Jiang","Jian Yao","Wei Liu"],"abstract":"Recently, caption generation with an encoder-decoder framework has been\nextensively studied and applied in different domains, such as image captioning,\ncode captioning, and so on. In this paper, we propose a novel architecture,\nnamely Auto-Reconstructor Network (ARNet), which, coupling with the\nconventional encoder-decoder framework, works in an end-to-end fashion to\ngenerate captions. ARNet aims at reconstructing the previous hidden state with\nthe present one, besides behaving as the input-dependent transition operator.\nTherefore, ARNet encourages the current hidden state to embed more information\nfrom the previous one, which can help regularize the transition dynamics of\nrecurrent neural networks (RNNs). Extensive experimental results show that our\nproposed ARNet boosts the performance over the existing encoder-decoder models\non both image captioning and source code captioning tasks. Additionally, ARNet\nremarkably reduces the discrepancy between training and inference processes for\ncaption generation. Furthermore, the performance on permuted sequential MNIST\ndemonstrates that ARNet can effectively regularize RNN, especially on modeling\nlong-term dependencies. Our code is available at:\nhttps://github.com/chenxinpeng/ARNet","url_abs":"http://arxiv.org/abs/1803.11439v2","url_pdf":"http://arxiv.org/pdf/1803.11439v2.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":"regularizing-rnns-for-caption-generation-by","repo_url":"https://github.com/chenxinpeng/ARNet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"caption-generation","task_name":"Caption Generation"},{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"image-captioning","task_name":"Image Captioning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.11439","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}