Michael Poli

Machine-learning researcher whose work spans structured and implicit data representations and efficient sequence architectures; lead author, with Stefano Ermon's group, of "Self-Similarity Priors: Neural Collages as Differentiable Fractal Representations" (NeurIPS 2022).

Matters to this vault as the researcher behind the single paper found that explicitly bridges two literatures this vault otherwise found citing past each other: fractal/IFS image compression and implicit neural representations. Poli et al.'s Neural Collages paper cites both canons in the same related-work section, reformulates Barnsley's Collage Theorem as a differentiable operator, and benchmarks the result against both fractal compression and COIN — the one document that turns a structural rhyme into a documented, if one-directional, lineage.

References

• claim-poli-et-al-2022-neural-collages-bridges-fractal-and-inr-canons · claim-fractal-compression-to-implicit-neural-representations-bridge
• entity-michael-barnsley · entity-collage-theorem