{"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/adding-gradient-noise-improves-learning-for","title":"Adding Gradient Noise Improves Learning for Very Deep Networks","arxiv_id":"1511.06807","date":"2015-11-21","proceeding":null,"authors":["Arvind Neelakantan","Luke Vilnis","Quoc V. Le","Ilya Sutskever","Lukasz Kaiser","Karol Kurach","James Martens"],"abstract":"Deep feedforward and recurrent networks have achieved impressive results in\nmany perception and language processing applications. This success is partially\nattributed to architectural innovations such as convolutional and long\nshort-term memory networks. The main motivation for these architectural\ninnovations is that they capture better domain knowledge, and importantly are\neasier to optimize than more basic architectures. Recently, more complex\narchitectures such as Neural Turing Machines and Memory Networks have been\nproposed for tasks including question answering and general computation,\ncreating a new set of optimization challenges. In this paper, we discuss a\nlow-overhead and easy-to-implement technique of adding gradient noise which we\nfind to be surprisingly effective when training these very deep architectures.\nThe technique not only helps to avoid overfitting, but also can result in lower\ntraining loss. This method alone allows a fully-connected 20-layer deep network\nto be trained with standard gradient descent, even starting from a poor\ninitialization. We see consistent improvements for many complex models,\nincluding a 72% relative reduction in error rate over a carefully-tuned\nbaseline on a challenging question-answering task, and a doubling of the number\nof accurate binary multiplication models learned across 7,000 random restarts.\nWe encourage further application of this technique to additional complex modern\narchitectures.","url_abs":"http://arxiv.org/abs/1511.06807v1","url_pdf":"http://arxiv.org/pdf/1511.06807v1.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":"adding-gradient-noise-improves-learning-for","repo_url":"https://github.com/Songlielie/DKT-Plus","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"adding-gradient-noise-improves-learning-for","repo_url":"https://github.com/WinnieHAN/dnhmm","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"torch","reach":{"status":"ok"}},{"paper_slug":"adding-gradient-noise-improves-learning-for","repo_url":"https://github.com/ketranm/neuralHMM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"torch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"adding-gradient-noise-improves-learning-for","repo_url":"https://github.com/sunk/qneurons","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"question-answering","task_name":"Question Answering"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1511.06807","atlas_url":"https://app.syntology.ai/?focus=1511.06807","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1511.06807"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/WinnieHAN/dnhmm","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Songlielie/DKT-Plus","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/sunk/qneurons","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ketranm/neuralHMM","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran_draft_wrong":1,"ran_honours":1},"by_repo_kind":{"listed":{"samples":2,"ran":2,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":2,"samples":[{"code_sha256_prefix":"71138bd13a12773c","entry":"add_gradient_noise","repo":"sunk/qneurons","repo_kind":"listed","path":"gradient_noise.py","file_url":"https://github.com/sunk/qneurons/blob/HEAD/gradient_noise.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"71138bd13a12773c"}},{"code_sha256_prefix":"fde7eb75fcfa9d2b","entry":"lr_schedule","repo":"sunk/qneurons","repo_kind":"listed","path":"runqact.py","file_url":"https://github.com/sunk/qneurons/blob/HEAD/runqact.py","link_basis":"first_harvest_node","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"well_formed","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"fde7eb75fcfa9d2b"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}