{"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/what-can-linearized-neural-networks-actually","title":"What can linearized neural networks actually say about generalization?","arxiv_id":"2106.06770","date":"2021-06-12","proceeding":"NeurIPS 2021 12","authors":["Guillermo Ortiz-Jiménez","Seyed-Mohsen Moosavi-Dezfooli","Pascal Frossard"],"abstract":"For certain infinitely-wide neural networks, the neural tangent kernel (NTK) theory fully characterizes generalization, but for the networks used in practice, the empirical NTK only provides a rough first-order approximation. Still, a growing body of work keeps leveraging this approximation to successfully analyze important deep learning phenomena and design algorithms for new applications. In our work, we provide strong empirical evidence to determine the practical validity of such approximation by conducting a systematic comparison of the behavior of different neural networks and their linear approximations on different tasks. We show that the linear approximations can indeed rank the learning complexity of certain tasks for neural networks, even when they achieve very different performances. However, in contrast to what was previously reported, we discover that neural networks do not always perform better than their kernel approximations, and reveal that the performance gap heavily depends on architecture, dataset size and training task. We discover that networks overfit to these tasks mostly due to the evolution of their kernel during training, thus, revealing a new type of implicit bias.","url_abs":"https://arxiv.org/abs/2106.06770v2","url_pdf":"https://arxiv.org/pdf/2106.06770v2.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":"what-can-linearized-neural-networks-actually","repo_url":"https://github.com/gortizji/linearized-networks","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"jax","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[],"methods":[{"method_slug":"ntk","method_name":"NTK"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2106.06770","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2106.06770"}},"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/gortizji/linearized-networks","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"unverified":4},"by_repo_kind":{"official":{"samples":4,"ran":0,"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":"a97ccf4556b4a717","entry":"binary_cross_entropy_loss_with_logits","repo":"gortizji/linearized-networks","repo_kind":"official","path":"utils/loss.py","file_url":"https://github.com/gortizji/linearized-networks/blob/HEAD/utils/loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"a97ccf4556b4a717"}},{"code_sha256_prefix":"58979551de01a0e4","entry":"mixed_derivative_nad_decomposition","repo":"gortizji/linearized-networks","repo_kind":"official","path":"neural_kernels/nads.py","file_url":"https://github.com/gortizji/linearized-networks/blob/HEAD/neural_kernels/nads.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"58979551de01a0e4"}},{"code_sha256_prefix":"82c402b4b47c433b","entry":"mse_loss","repo":"gortizji/linearized-networks","repo_kind":"official","path":"utils/loss.py","file_url":"https://github.com/gortizji/linearized-networks/blob/HEAD/utils/loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"82c402b4b47c433b"}},{"code_sha256_prefix":"581398948f76cfae","entry":"softmax_cross_entropy_loss","repo":"gortizji/linearized-networks","repo_kind":"official","path":"utils/loss.py","file_url":"https://github.com/gortizji/linearized-networks/blob/HEAD/utils/loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"581398948f76cfae"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}