An Entropy-Based Graph Construction Method for Representing and Clustering Biological Data

Unsupervised learning methods are commonly used to perform the non-trivial task of uncovering structure in biological data. However, conventional approaches rely on methods that make assumptions about data distribution and reduce the dimensionality of the input data. Here we propose the incorporation of entropy related measures into the process of constructing graph-based representations for biological datasets in order to uncover their inner structure. Experimental results demonstrated the potential of the proposed entropy-based graph data representation to cope with biological applications related to unsupervised learning problems, such as metagenomic binning and neuronal spike sorting, in which it is necessary to organize data into unknown and meaningful groups. © 2020, Springer Nature Switzerland AG.

• Fecha de publicación: 2020
• Institución: Universidad de Medellín
• Repositorio: Repositorio UDEM
• Idioma: eng
• OAI Identifier: oai:repository.udem.edu.co:11407/5648
• Acceso en línea: http://hdl.handle.net/11407/5648
• Palabra clave: Biological data; Clustering; Entropy; Graph; Metagenomic binning; Spike sorting; Biomedical engineering; Biophysics; Entropy; Graphic methods; Machine learning; Unsupervised learning; Biological data; Clustering; Graph; Metagenomic binning; Spike-sorting; Sorting
• License: http://purl.org/coar/access_right/c_16ec