Constructive interpolation and generalization rates for neural ODEs: a control perspective

Antonio Álvarez-López, Lorenzo Liverani, Enrique Zuazua · 2026

Read Constructive interpolation and generalization rates for neural ODEs: a control perspective by Antonio Álvarez-López, Lorenzo Liverani, Enrique Zuazua (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-00469.txt.

Description

Álvarez and co-authors view neural ODE learning through the lens of control math. Such models let a net's state flow by a law that shifts with time. Fit and broad skill for them lacked crisp bounds in the past. The team study a class with fixed weights and clear time ties.

They prove this class can fit set data in an exact way. A stronger trait lets its flows send split cells to tiny target balls at will. That trait turns fit skill into broad skill by a plain route. Risk bounds match those of bin and near-kin tools when width scales with draws. Time ties prove key: two-layer time-free forms can fit but lack the strong trait.

Catalog record

CREATORAntonio Álvarez-López, Lorenzo Liverani, Enrique Zuazua
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE123 KB
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2606.00469

Browse the full library · About the Tower of Babel