Asier Gutiérrez-Fandiño

First author of "Persistent Homology Captures the Generalization of Neural Networks Without A Validation Set" (arXiv:2106.00012, 2021, with David Pérez-Fernández, Jordi Armengol-Estapé & Marta Villegas) — the second of the vault's two founding neural-network persistent-homology diagnostics, tracking persistence-diagram distance between successive training states as a validation-set-free correlate of generalization. Matters to the vault because his method's specific design choice — building the simplicial complex from every neuron and connection, with no subsampling step at all — turned out to be structurally immune, by construction, to the hub-selection sampling artifact the vault's Byzantine-trade-network study documents elsewhere.

References

• claim-gutierrez-fandino-2021-persistence-diagram-distance-tracks-generalization · claim-gutierrez-fandino-2021-method-subsamples-nothing-uses-full-network
• observation-persistent-homology-gradient-free-bridge-widrow-byzantine · observation-hub-selection-artifact-absent-by-design-in-founding-ph-generalization-papers
• entity-tolga-birdal · question-tda-neural-net-sampling-artifact-risk