@inproceedings{rangzan2025mixture,
    type = {inproceedings},
    key = {rangzan2025mixture},
    title = {Mixture of Geographical Experts: Disentangling Earth},
    author = {Moien Rangzan and Gregory Duveiller and Maha Shadaydeh and Markus Reichstein and Joachim Denzler},
    booktitle = {EurIPS Workshop on Advances in Representation Learning for Earth Observation (EurIPS-WS)},
    year = {2025},
    url = {https://sites.google.com/view/reoeurips},
    note = {},
    abstract = {Most domain generalization techniques assume there exists a stable predictive relationship from input features to labels across domains, an assumption that breaks in many Earth observation tasks, where stable signals are weak or geographically confined. We introduce a sparse, geo-routed Mixture of Geographical Experts (MoGE) that explicitly disentangles global invariance from spatial variation. A shared invariant expert captures features that hold everywhere, while metadata-driven routing activates a subset of geo-specialized experts forced to learn region-specific cues. This separation lets experts self-organize into continuous, concept-consistent regions, discovering domains rather than handcrafting them, while the invariant path remains robust across space. MoGE's factorization yields strong performance on generalization benchmarks.},
}