Are phylogenetic reconstruction methods considered machine-learning approaches?

Question

[fhsantanna]

Could bayesian analysis, maximum likelihood, neighbour joining be considered unsupervised machine-learning approaches?

Answer 1

[Jean-Karim Heriche]

Definitions for machine learning vary. If you pick some definition around learning from data then you could consider many data analysis approaches as machine learning. Sometimes, it's presented as the science to automatically build models from data. However, in practice, the term machine learning is often used to refer to statistical methods focusing on finding patterns and making prediction (i.e. classification and regression for the supervised case and clustering for the unsupervised case). So the context seems to play a role. Building a phylogenetic tree is a form of hierarchical clustering so could be considered machine learning by most definitions except that the people doing phylogenetics do not come from the traditional machine learning community so they don't call their activity as such.

[fhsantanna]

Thank you very much. As I thought, it seems only a nomenclature divergence for the same thing.