Bergomi M. G., Frosini P., Giorgi D., Quercioli N.
55N35 (Primary) Machine Learning (stat.ML) Mathematics - Algebraic Topology topological data analysis FOS: Mathematics Computer Vision and Pattern Recognition Computer Networks and Communications 68U05 Machine Learning 65D18 (Secondary) Computer Vision and Pattern Recognition (cs.CV) FOS: Computer and information sciences Equivariant operators 57S10 Topological Data Analysis persistent homology 47H09 natural pseudo-distance Machine Learning (cs.LG) Mathematics - Operator Algebras Computer Science - Machine Learning Statistics - Machine Learning persistent topology Artificial Intelligence Shape classification Operator Algebras (math.OA) Algebraic Topology (math.AT) Human-Computer Interaction 54H15 Group equivariant non-expansive operator Software Metric learning Computer Science - Computer Vision and Pattern Recognition
We provide a general mathematical framework for group and set equivariance in machine learning. We define group equivariant non-expansive operators (GENEOs) as maps between function spaces associated with groups of transformations. We study the topological and metric properties of the space of GENEOs to evaluate their approximating power and set the basis for general strategies to initialize and compose operators. We define suitable pseudo-metrics for the function spaces, the equivariance groups and the set of non-expansive operators. We prove that, under suitable assumptions, the space of GENEOs is compact and convex. These results provide fundamental guarantees in a machine learning perspective. By considering isometry-equivariant non-expansive operators, we describe a simple strategy to select and sample operators. Thereafter, we show how selected and sampled operators can be used both to perform classical metric learning and to inject knowledge in artificial neural networks.
Source: Nature Machine Intelligence 1 (2019): 423–433. doi:10.1038/s42256-019-0087-3
Publisher: Springer Nature
@article{oai:it.cnr:prodotti:408494, title = {Towards a topological-geometrical theory of group equivariant non-expansive operators for data analysis and machine learning}, author = {Bergomi M. G. and Frosini P. and Giorgi D. and Quercioli N.}, publisher = {Springer Nature}, doi = {10.1038/s42256-019-0087-3 and 10.48550/arxiv.1812.11832}, journal = {Nature Machine Intelligence}, volume = {1}, pages = {423–433}, year = {2019} }