{"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/ablation-studies-in-artificial-neural","title":"Ablation Studies in Artificial Neural Networks","arxiv_id":"1901.08644","date":"2019-01-24","proceeding":null,"authors":["Richard Meyes","Melanie Lu","Constantin Waubert de Puiseau","Tobias Meisen"],"abstract":"Ablation studies have been widely used in the field of neuroscience to tackle\ncomplex biological systems such as the extensively studied Drosophila central\nnervous system, the vertebrate brain and more interestingly and most\ndelicately, the human brain. In the past, these kinds of studies were utilized\nto uncover structure and organization in the brain, i.e. a mapping of features\ninherent to external stimuli onto different areas of the neocortex. considering\nthe growth in size and complexity of state-of-the-art artificial neural\nnetworks (ANNs) and the corresponding growth in complexity of the tasks that\nare tackled by these networks, the question arises whether ablation studies may\nbe used to investigate these networks for a similar organization of their inner\nrepresentations. In this paper, we address this question and performed two\nablation studies in two fundamentally different ANNs to investigate their inner\nrepresentations of two well-known benchmark datasets from the computer vision\ndomain. We found that features distinct to the local and global structure of\nthe data are selectively represented in specific parts of the network.\nFurthermore, some of these representations are redundant, awarding the network\na certain robustness to structural damages. We further determined the\nimportance of specific parts of the network for the classification task solely\nbased on the weight structure of single units. Finally, we examined the ability\nof damaged networks to recover from the consequences of ablations by means of\nrecovery training. We argue that ablations studies are a feasible method to\ninvestigate knowledge representations in ANNs and are especially helpful to\nexamine a networks robustness to structural damages, a feature of ANNs that\nwill become increasingly important for future safety-critical applications.","url_abs":"http://arxiv.org/abs/1901.08644v2","url_pdf":"http://arxiv.org/pdf/1901.08644v2.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":"ablation-studies-in-artificial-neural","repo_url":"https://github.com/RichardMeyes/AblationStudies","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1901.08644","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1901.08644"}},"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/RichardMeyes/AblationStudies","reach":null}],"summary":{"ran_fixture":1},"by_repo_kind":{"official":{"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":"8858df296bc74b62","entry":"calc_unit_struct_metric","repo":"RichardMeyes/AblationStudies","repo_kind":"official","path":"code/plot_custom_knockouts_l1.py","file_url":"https://github.com/RichardMeyes/AblationStudies/blob/HEAD/code/plot_custom_knockouts_l1.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"BSD-3-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"8858df296bc74b62"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}