J. L. Sainz-Pardo Auñon
Neighbors Classification are non-parametric supervised learning methods that assigns a class to a data point based on the classes of its closest neighbors in a feature space. The most common techniques based on Neighbors Classification are k-Nearest Neighbors Classification (k-NN) and Radius Neighbors Classification (R-NN).
Prototype selection is a technique used in machine learning, to reduce the size of datasets without losing classification accuracy when Neighbors Classification methods are employed. Among the main techniques of prototype selection are Condensed Nearest Neighbor (CNN), Edited Nearest Neighbor (ENN), and others.
This work presents different novel methods for prototype selection based on integer programming models, addressing both their exact resolution and their resolution using heuristic methods provided. A wide computational experience is addressed by comparing the proposed methods with each other and with the classical ones.
Palabras clave: Prototype selection, integer programming, nearest neighbor classification, machine learning
Programado
Clasificación y reconocimiento de patrones
11 de junio de 2025 15:30
Auditorio 1. Ricard Vinyes