Geometric Mean Pooling for Equal-Weight Multiplicative Coarse-Graining
Abstract
As an alternative to the additive and extremal biases of average and max pooling, we introduce Geometric Mean Pooling (GMP), a signed pooling operator that combines the product of feature signs with the geometric mean of feature magnitudes. Motivated by local-to-global composition in quantum many-body physics, GMP retains both joint sign information and a characteristic multiplicative scale without introducing learnable pooling parameters. We show that non-overlapping hierarchical GMP preserves the corresponding global multiplicative statistic and evaluate it on synthetic sequence tasks, iterative coarse-graining, image classification, and molecular lipophilicity regression. On the synthetic tasks, GMP recovers product-based signals more accurately than average and max pooling and maintains predictive performance under the tested levels of multiplicative input noise. On image and molecular data, however, its effectiveness depends on the representation, target parameterization, and placement of local and global pooling. These results position GMP as a complementary, regime-dependent inductive bias for tasks in which equal-weight multiplicative composition is plausible, rather than as a universal replacement for standard pooling operators.
Related
- Commutative Coarse-graining on the Sphere
- Coarse-Graining Optimal Transport for Rollout-Free Post-Training of Flow Matching
- Conditional preservation of chimera states under equitable network coarse-graining
- Initial-Trace Insufficiency and Entropy-Accounting Separation in Finite-Volume Coarse Graining
- Geometric difference families and related variable-weight geometric orthogonal codes
BibTeX
@misc{wu2026geometric,
title = {{Geometric Mean Pooling for Equal-Weight Multiplicative Coarse-Graining}},
author = {Ang-Kun Wu and Fangdi Wen and Jingtao Zhang},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.21876},
eprint = {2609.21876},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.21876},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Geometric Mean Pooling for Equal-Weight Multiplicative Coarse-Graining
AU - Ang-Kun Wu
AU - Fangdi Wen
AU - Jingtao Zhang
PY - 2026
DA - 2026/09/25
PB - arXiv
DO - 10.48550/arxiv.2609.21876
UR - https://arxiv.org/abs/2609.21876
AB - As an alternative to the additive and extremal biases of average and max pooling, we introduce Geometric Mean Pooling (GMP), a signed pooling operator that combines the product of feature signs with the geometric mean of feature magnitudes. Motivated by local-to-global composition in quantum many-body physics, GMP retains both joint sign information and a characteristic multiplicative scale without introducing learnable pooling parameters. We show that non-overlapping hierarchical GMP preserves the corresponding global multiplicative statistic and evaluate it on synthetic sequence tasks, iterative coarse-graining, image classification, and molecular lipophilicity regression. On the synthetic tasks, GMP recovers product-based signals more accurately than average and max pooling and maintains predictive performance under the tested levels of multiplicative input noise. On image and molecular data, however, its effectiveness depends on the representation, target parameterization, and placement of local and global pooling. These results position GMP as a complementary, regime-dependent inductive bias for tasks in which equal-weight multiplicative composition is plausible, rather than as a universal replacement for standard pooling operators.
ER -