Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 3 min ago
PreprintarXiv licencearXiv

A Local Geometric Analysis of Maximal Coding Rate Reduction via Error Bounds

  • Peng Wang
  • Po Chen
  • Huikang Liu
  • Rujun Jiang
Open at sourcePDFDOICite

Abstract

While the maximal coding rate reduction (MCR2) objective has become a useful principle for learning compact and discriminative representations, a theoretical understanding of its optimization geometry remains limited, especially regarding why first-order methods often exhibit fast convergence near its local maximizers. In this work, we show that the MCR2 objective satisfies an error-bound property in a neighborhood of its local maximizers. This property guarantees that the distance to the local maximizer set can be bounded above by the gradient norm at any point in the neighborhood, directly implying regularity conditions such as the Polyak--Łojasiewicz inequality and quadratic growth. Leveraging this property, we prove that first-order methods converge linearly to a local maximizer of the MCR2 objective under mild conditions. Numerical experiments validate the predicted local convergence behavior and provide empirical evidence that optimizing the MCR2 objective improves representation geometry while maintaining competitive downstream accuracy.

Related

  1. Entropy and Distributed Source Coding of Connected Soft Random Geometric Graphs
    Oliver Baker, Carl P. Dettmann
    arXivPreprintCC BY 4.0Full textProbabilityarXiv
  2. Maximal geometric quotients of flag varieties
    Lorenzo Barban, Michele Bianco, Luis E. Solá Conde
    arXivPreprintCC BY-SA 4.0Full textAlgebra & number theoryarXiv
  3. Improved Lp bounds for the helical maximal function in dimensions n ≥ 5
    Changkeun Oh, Jaehyun Woo
    arXivPreprintCC0Full textAnalysis & PDEsarXiv
  4. CAS I: A Geometric Coding Theorem
    Romie Banerjee
    arXivPreprintCC BY 4.0Full textAlgebra & number theoryarXiv
  5. Improved weighted bounds for the strong maximal function
    Sheldy Ombrosi, Guillermo Rey
    arXivPreprintCC BY 4.0Full textAnalysis & PDEsarXiv

Cite

BibTeXRIS
BibTeX
@misc{wang2026local,
  title = {{A Local Geometric Analysis of Maximal Coding Rate Reduction via Error Bounds}},
  author = {Peng Wang and Po Chen and Huikang Liu and Rujun Jiang},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.05072},
  eprint = {2610.05072},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.05072},
}
RIS
TY  - UNPB
TI  - A Local Geometric Analysis of Maximal Coding Rate Reduction via Error Bounds
AU  - Peng Wang
AU  - Po Chen
AU  - Huikang Liu
AU  - Rujun Jiang
PY  - 2026
DA  - 2026/10/04
PB  - arXiv
DO  - 10.48550/arxiv.2610.05072
UR  - https://arxiv.org/abs/2610.05072
AB  - While the maximal coding rate reduction (MCR^2) objective has become a useful principle for learning compact and discriminative representations, a theoretical understanding of its optimization geometry remains limited, especially regarding why first-order methods often exhibit fast convergence near its local maximizers. In this work, we show that the MCR^2 objective satisfies an error-bound property in a neighborhood of its local maximizers. This property guarantees that the distance to the local maximizer set can be bounded above by the gradient norm at any point in the neighborhood, directly implying regularity conditions such as the Polyak--Łojasiewicz inequality and quadratic growth. Leveraging this property, we prove that first-order methods converge linearly to a local maximizer of the MCR^2 objective under mild conditions. Numerical experiments validate the predicted local convergence behavior and provide empirical evidence that optimizing the MCR^2 objective improves representation geometry while maintaining competitive downstream accuracy.
ER  -