{"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/self-normalizing-neural-networks","title":"Self-Normalizing Neural Networks","arxiv_id":"1706.02515","date":"2017-06-08","proceeding":"NeurIPS 2017 12","authors":["Günter Klambauer","Thomas Unterthiner","Andreas Mayr","Sepp Hochreiter"],"abstract":"Deep Learning has revolutionized vision via convolutional neural networks\n(CNNs) and natural language processing via recurrent neural networks (RNNs).\nHowever, success stories of Deep Learning with standard feed-forward neural\nnetworks (FNNs) are rare. FNNs that perform well are typically shallow and,\ntherefore cannot exploit many levels of abstract representations. We introduce\nself-normalizing neural networks (SNNs) to enable high-level abstract\nrepresentations. While batch normalization requires explicit normalization,\nneuron activations of SNNs automatically converge towards zero mean and unit\nvariance. The activation function of SNNs are \"scaled exponential linear units\"\n(SELUs), which induce self-normalizing properties. Using the Banach fixed-point\ntheorem, we prove that activations close to zero mean and unit variance that\nare propagated through many network layers will converge towards zero mean and\nunit variance -- even under the presence of noise and perturbations. This\nconvergence property of SNNs allows to (1) train deep networks with many\nlayers, (2) employ strong regularization, and (3) to make learning highly\nrobust. Furthermore, for activations not close to unit variance, we prove an\nupper and lower bound on the variance, thus, vanishing and exploding gradients\nare impossible. We compared SNNs on (a) 121 tasks from the UCI machine learning\nrepository, on (b) drug discovery benchmarks, and on (c) astronomy tasks with\nstandard FNNs and other machine learning methods such as random forests and\nsupport vector machines. SNNs significantly outperformed all competing FNN\nmethods at 121 UCI tasks, outperformed all competing methods at the Tox21\ndataset, and set a new record at an astronomy data set. The winning SNN\narchitectures are often very deep. Implementations are available at:\ngithub.com/bioinf-jku/SNNs.","url_abs":"http://arxiv.org/abs/1706.02515v5","url_pdf":"http://arxiv.org/pdf/1706.02515v5.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":"self-normalizing-neural-networks","repo_url":"https://github.com/bioinf-jku/SNNs","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/BenjiKCF/SNN-with-embeddings-for-Malware-Prediction","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/BioImageInformatics/tfmodels","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/blackPacha/VAE_ABIDE1","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/dmbernaal/Daedalus","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/domschl/syncognite","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/jonnor/datascience-master","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/mshunshin/SegNetCMR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/omrijsharon/NetworkModule","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/2023-MindSpore-4/Code11/tree/main/snn_mlp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/2023-MindSpore-4/Code7/tree/main/snn_mlp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/MaximeVandegar/Papers-in-100-Lines-of-Code/tree/main/Self_Normalizing_Neural_Networks","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"self-normalizing-neural-networks","repo_url":"https://github.com/Mind23-2/MindCode-88/tree/main/snn_mlp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":"astronomy","task_name":"Astronomy"},{"task_slug":"machine-learning","task_name":"BIG-bench Machine Learning"},{"task_slug":"drug-discovery","task_name":"Drug Discovery"},{"task_slug":"pulsar-prediction","task_name":"Pulsar Prediction"}],"methods":[{"method_slug":"batch-normalization","method_name":"Batch Normalization"}],"datasets_introduced":[],"methods_introduced":[{"slug":"snn","name":"SNN","full_name":"Spiking Neural Networks"}],"results":[{"leaderboard":"/sota/drug-discovery-on-tox21","task":"Drug Discovery","dataset":"Tox21","model":"SNN (SELU Network)","rank_in_archive_order":8,"of":11,"metrics":{"AUC":"0.845"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1706.02515","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1706.02515"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+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/bioinf-jku/SNNs","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/blackPacha/VAE_ABIDE1","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/BenjiKCF/SNN-with-embeddings-for-Malware-Prediction","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/dmbernaal/Daedalus","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/2023-MindSpore-4/Code11/tree/main/snn_mlp","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/MaximeVandegar/Papers-in-100-Lines-of-Code/tree/main/Self_Normalizing_Neural_Networks","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/domschl/syncognite","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/BioImageInformatics/tfmodels","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Mind23-2/MindCode-88/tree/main/snn_mlp","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/mshunshin/SegNetCMR","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/omrijsharon/NetworkModule","reach":{"status":"unanswered"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/jonnor/datascience-master","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/2023-MindSpore-4/Code7/tree/main/snn_mlp","reach":null}],"summary":{"ran_honours":1},"by_repo_kind":{"listed":{"samples":1,"ran":1,"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":0,"samples":[{"code_sha256_prefix":"cd177ac59dc1b42c","entry":"train","repo":"MaximeVandegar/Papers-in-100-Lines-of-Code","repo_kind":"listed","path":"Self_Normalizing_Neural_Networks/selu.py","file_url":"https://github.com/MaximeVandegar/Papers-in-100-Lines-of-Code/blob/HEAD/Self_Normalizing_Neural_Networks/selu.py","link_basis":"first_harvest_node","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"cd177ac59dc1b42c"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}