A topological framework tracks how data shape (clusters, loops, holes) evolves across the layers of a multilayer perceptron, with stability guarantees and a practical approximation algorithm.
Dey, Facundo Mémoli, and Yusu Wang
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Latent Space Topology Evolution in Multilayer Perceptrons
A topological framework tracks how data shape (clusters, loops, holes) evolves across the layers of a multilayer perceptron, with stability guarantees and a practical approximation algorithm.