On Convolutions, Intrinsic Dimension, and Diffusion Models
Kin Kwan Leung [email protected], Rasa Hosseinzadeh [email protected], Gabriel Loaiza-Ganem [email protected] · 2025
Read On Convolutions, Intrinsic Dimension, and Diffusion Models by Kin Kwan Leung [email protected], Rasa Hosseinzadeh [email protected], Gabriel Loaiza-Ganem [email protected] (2025) 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-2506-20705.txt.
Description
Kin Kwan Leung, Rasa Hosseinzadeh, and Gabriel Loaiza-Ganem link convolutions, intrinsic dimension, and diffusion models in two thousand twenty-five. The manifold view says real data lie on thin sheets in big spaces. Diffusion models add noise by convolution, then learn to undo it. A past estimator reads local dimension from how density shifts with noise.
That estimator, FLIPD, was proved only for flat sheet data. The team proves it under far more real assumptions. A twin result holds when uniform noise swaps in for Gaussian noise. Local dimension gauges data complexity and flags outliers and fakes. The proof firms up a tool in wide use.
Catalog record
| CREATOR | Kin Kwan Leung [email protected], Rasa Hosseinzadeh [email protected], Gabriel Loaiza-Ganem [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 70 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-20705