A topological description of loss surfaces based on Betti Numbers

Published: 01 Jan 2024, Last Modified: 01 Aug 2025Neural Networks 2024EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Establishment of bounds on the topological complexity of loss functions in terms of Betti numbers’s characterization.•Bounds are provided for Mean Squared Error and Binary Cross Entropy loss functions and for networks with Pfaffian activation functions.•Analysis of the effect of skip connections and ℓ2<math><msub is="true"><mrow is="true"><mi is="true">ℓ</mi></mrow><mrow is="true"><mn is="true">2</mn></mrow></msub></math> regularization on the topological complexity.
Loading