EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Top-down and bottom-up attention for joint pattern classification and reconstruction

Ricardo A Veiga iD, Bonifacio Silvano Zanutto

PLOS One · 01.07.2026 · 21 (7) · e0351985

DOI: 10.1371/journal.pone.0351985 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS One, online erschienen am 01.07.2026. Verfasst von Ricardo A Veiga und 1 weiteren Autor:innen. Beteiligte Einrichtung: Facultad de Ingeniería, Universidad de Buenos Aires, Instituto de Ingeniería Biomédica, Buenos Aires, Argentina. Hauptthemen laut MEDLINE: Pattern Recognition, Automated. Die Arbeit nennt 37 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Secretaría de Ciencia y Técnica, Universidad de Buenos Aires.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

We introduce a recurrent inference framework for the Classification and Reconstruction of Overlapping Patterns (CROP) in mixtures formed by overlapping two patterns drawn from the same distribution. The framework alternates between bottom-up classification and top-down generative reconstruction within an iterative inference procedure. At each iteration, the method estimates the most likely class present in the mixture, reconstructs the corresponding signal using a conditional generative model, and applies a mask to isolate that component. This classification-guided reconstruction progressively separates the overlapping signals while also producing their class labels. The objective is therefore to iteratively separate and classify the overlapping patterns rather than perform general blind source separation. An important feature of the framework is that the generative model can be trained using only clean samples, without requiring paired mixed–clean training data. The iterative procedure implicitly implements a form of attention in which saliency- and priority-driven estimates guide the masking and reconstruction of individual patterns. Experimental results on mixtures of handwritten digits show that the proposed framework can successfully separate and classify overlapping patterns through this iterative classification–reconstruction process.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Pattern Recognition, AutomatedAlgorithmsClassification AlgorithmsHumans

Autor:innen und Institutionen

  • Facultad de Ingeniería, Universidad de Buenos Aires, Instituto de Ingeniería Biomédica, Buenos Aires, Argentina.
  • Laboratorio de Biología del Comportamiento, Instituto de Biología y Medicina Experimental (IBYME-CONICET), Buenos Aires, Argentina.

ORCID-Profile: Ricardo A Veiga

Förderung

  • Secretaría de Ciencia y Técnica, Universidad de Buenos Aires

Zitieren

APA

Veiga, R. A., & Zanutto, B. S. (2026). Top-down and bottom-up attention for joint pattern classification and reconstruction. PLOS One, 21(7), e0351985. https://doi.org/10.1371/journal.pone.0351985

Vancouver

Veiga RA, Zanutto BS. Top-down and bottom-up attention for joint pattern classification and reconstruction. PLOS One. 2026;21(7):e0351985. doi:10.1371/journal.pone.0351985

BibTeX
@article{Veiga2026,
  author = {Veiga, Ricardo A and Zanutto, Bonifacio Silvano},
  title = {Top-down and bottom-up attention for joint pattern classification and reconstruction},
  journal = {PLOS One},
  year = {2026},
  volume = {21},
  number = {7},
  pages = {e0351985},
  doi = {10.1371/journal.pone.0351985}
}
Kostenlos starten →