{
  "abstract": "Learning from imperfect data is a central theme in machine learning, connecting practical questions of robustness to fundamental questions of learnability. Here, we examine attribute noise: learning from corrupted inputs while keeping the labels intact, a setting that has received considerably less analytical attention than its label-noise counterpart. We consider two types of corruption models: additive noise and replacement noise. Through experiments with multilayer perceptrons (MLPs) on corrupted classification datasets, we find that neural networks remain robust, maintaining well-above-chance accuracy even when inputs are >90% corrupted—far beyond human recognition. To understand this robustness, we analyze infinite-width networks in the heavy-corruption regime using a mean-field-inspired approach and derive a leading-order decision rule for the classification outcome: The network implements a prototype rule, the nearest-class-mean, assigning each test point to the class whose training-set average it most closely resembles. This leading-order decision rule is universal across a broad range of MLP architectures, holding for any depth, as well as a wide class of activation functions and noise distributions. The same centroid mechanism closely matches finite-width network behavior in our experiments and provides an interpretable and analytically tractable account of why learning can succeed even when individual training examples carry almost no signal.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": "article",
  "arxiv": null,
  "author_total": 4,
  "authors": [
    {
      "name": "Justin Tahmassebpur"
    },
    {
      "name": "Asadullah Bhuiyan"
    },
    {
      "name": "Hyejin Kim"
    },
    {
      "name": "Omri Lesser"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/8wysa7vj0j5h7.bib",
  "cited_by": 0,
  "doi": "10.1103/qkrg-wbbj",
  "fulltext_url": "http://doi.org/10.1103/qkrg-wbbj",
  "id": "8wysa7vj0j5h7",
  "issns": [
    "2643-1564"
  ],
  "landing_url": "http://link.aps.org/doi/10.1103/qkrg-wbbj",
  "language": "en",
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "journal",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [
    [
      "feed-journal",
      "physics/general"
    ],
    [
      "lcc",
      "QC1-999"
    ],
    [
      "openalex",
      "1702|Neural Networks and Applications"
    ]
  ],
  "openalex": "W7164392250",
  "pdf_url": "http://doi.org/10.1103/qkrg-wbbj",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-23",
  "publisher": "American Physical Society",
  "ris": "https://research.draconitic.com/paper/8wysa7vj0j5h7.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:17:18Z",
      "native_id": "10.1103/qkrg-wbbj",
      "source": "feed-aps",
      "url": "http://link.aps.org/doi/10.1103/qkrg-wbbj"
    },
    {
      "first_seen": "2026-10-30T13:44:16Z",
      "native_id": "81687ebe9c3f447a83bcc2d177b6b042",
      "source": "doaj",
      "url": "https://doi.org/10.1103/qkrg-wbbj"
    }
  ],
  "subjects": [
    {
      "field": "physics",
      "origin": "journal-openalex:3107|Quantum many-body systems",
      "subfield": "amo"
    }
  ],
  "title": "Neural networks become centroid classifiers when trained on heavily corrupted inputs",
  "title_text": "Neural networks become centroid classifiers when trained on heavily corrupted inputs",
  "type": "peer-reviewed",
  "updated_at": "2026-10-30T13:44:16Z",
  "url": "https://research.draconitic.com/paper/8wysa7vj0j5h7",
  "venue": "Physical Review Research",
  "versions": []
}