{
  "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 multi-layer 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": null,
  "arxiv": "2606.11319",
  "author_total": 4,
  "authors": [
    {
      "name": "Justin Tahmassebpur"
    },
    {
      "name": "Asadullah Bhuiyan"
    },
    {
      "name": "Hyejin Kim"
    },
    {
      "name": "Omri Lesser"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/b1619br3f4yzv.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2606.11319",
  "fulltext_url": null,
  "id": "b1619br3f4yzv",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2606.11319",
  "language": null,
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "article",
  "licence_url": "http://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cond-mat.dis-nn"
    ],
    [
      "arxiv",
      "cs.LG"
    ]
  ],
  "openalex": null,
  "pdf_url": "https://arxiv.org/pdf/2606.11319",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-22",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/b1619br3f4yzv.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:35:15Z",
      "native_id": "2606.11319",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2606.11319"
    }
  ],
  "subjects": [
    {
      "field": "physics",
      "origin": "arxiv:cond-mat.dis-nn",
      "subfield": "condensed-matter"
    }
  ],
  "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": "preprint",
  "updated_at": "2026-11-05T15:20:55Z",
  "url": "https://research.draconitic.com/paper/b1619br3f4yzv",
  "venue": "arXiv",
  "versions": []
}