arXiv:2511.17427v1 Announce Type: new 
Abstract: We present a differentiable extension of the VEROS ocean model, enabling automatic differentiation through its dynamical core. We describe the key modifications required to make the model fully compatible with JAX autodifferentiation framework and evaluate the numerical consistency of the resulting implementation. Two illustrative applications are then demonstrated: (i) the correction of an initial ocean state through gradient-based optimization, and (ii) the calibration of unknown physical parameters directly from model observations. These examples highlight how differentiable programming can facilitate end-to-end learning and parameter tuning in ocean modeling. Our implementation is available online.

تم تقديم امتداد تفاضلي جديد لنموذج المحيطات VEROS، مما يتيح التفاضل التلقائي من خلال نواته الديناميكية. يعزز هذا التطور توافق النموذج مع إطار التفاضل التلقائي JAX ويتضمن تعديلات رئيسية لضمان الاتساق العددي. يتم عرض تطبيقين يبرزان فائدته: تصحيح حالة المحيط الأولية من خلال تحسين قائم على التدرجات، ومعايرة المعلمات الفيزيائية غير المعروفة مباشرة من ملاحظات النموذج.

Se ha introducido una nueva extensión diferenciable del modelo oceánico VEROS, que permite la diferenciación automática a través de su núcleo dinámico. Este desarrollo mejora la compatibilidad del modelo con el marco de autodiferenciación JAX e incluye modificaciones clave para garantizar la consistencia numérica. Dos aplicaciones demuestran su utilidad: la optimización de estados oceánicos iniciales y la calibración de parámetros físicos a partir de observaciones del modelo.

Une nouvelle extension différentiable du modèle océanique VEROS a été introduite, permettant la différentiation automatique à travers son noyau dynamique. Ce développement améliore la compatibilité du modèle avec le cadre de différentiation automatique JAX et comprend des modifications clés pour garantir la cohérence numérique. Deux applications démontrent son utilité : l'optimisation des états océaniques initiaux et la calibration des paramètres physiques à partir des observations du modèle.

A new differentiable extension of the VEROS ocean model has been introduced, allowing for automatic differentiation through its dynamical core. This development enhances the model's compatibility with the JAX autodifferentiation framework and includes key modifications to ensure numerical consistency. Two applications demonstrate its utility: optimizing initial ocean states and calibrating physical parameters from model observations.

Towards fully differentiable neural ocean model with Veros

arXiv:2511.18157v1 Announce Type: new 
Abstract: Three-dimensional rigid-body transforms, i.e. rotations and translations, are central to modern differentiable machine learning pipelines in robotics, vision, and simulation. However, numerically robust and mathematically correct implementations, particularly on SO(3), are error-prone due to issues such as axis conventions, normalizations, composition consistency and subtle errors that only appear in edge cases. SciPy's spatial.transform module is a rigorously tested Python implementation. However, it historically only supported NumPy, limiting adoption in GPU-accelerated and autodiff-based workflows. We present a complete overhaul of SciPy's spatial.transform functionality that makes it compatible with any array library implementing the Python array API, including JAX, PyTorch, and CuPy. The revised implementation preserves the established SciPy interface while enabling GPU/TPU execution, JIT compilation, vectorized batching, and differentiation via native autodiff of the chosen backend. We demonstrate how this foundation supports differentiable scientific computing through two case studies: (i) scalability of 3D transforms and rotations and (ii) a JAX drone simulation that leverages SciPy's Rotation for accurate integration of rotational dynamics. Our contributions have been merged into SciPy main and will ship in the next release, providing a framework-agnostic, production-grade basis for 3D spatial math in differentiable systems and ML.

أعلنت مكتبة SciPy عن تحديث كبير لوحدة spatial.transform الخاصة بها، والتي تدعم الآن التحويلات ثلاثية الأبعاد القابلة للاشتقاق والمتوافقة مع مكتبات المصفوفات المختلفة، بما في ذلك JAX وPyTorch وCuPy. تعالج هذه التحديثات القيود السابقة المتعلقة بتسريع GPU والتفاضل التلقائي، مما يعزز قابليتها للتطبيق في سير العمل الخاص بالتعلم الآلي.

La biblioteca SciPy ha anunciado una actualización significativa de su módulo spatial.transform, que ahora admite transformaciones 3D diferenciables compatibles con varias bibliotecas de arreglos, incluidas JAX, PyTorch y CuPy. Esta revisión aborda las limitaciones anteriores relacionadas con la aceleración por GPU y la diferenciación automática, mejorando su aplicabilidad en flujos de trabajo de aprendizaje automático.

La bibliothèque SciPy a annoncé une mise à jour significative de son module spatial.transform, qui prend désormais en charge les transformations 3D différentiables compatibles avec diverses bibliothèques de tableaux, y compris JAX, PyTorch et CuPy. Cette refonte répond aux limitations précédentes liées à l'accélération GPU et à la différenciation automatique, améliorant son applicabilité dans les flux de travail d'apprentissage automatique.

The SciPy library has announced a significant update to its spatial.transform module, which now supports differentiable 3D transformations compatible with various array libraries, including JAX, PyTorch, and CuPy. This overhaul addresses previous limitations related to GPU acceleration and automatic differentiation, enhancing its applicability in machine learning workflows.

Towards fully differentiable neural ocean model with Veros

Was this article worth reading? Share it

Dyad

Attentive AI

Cometapi-e0d0fd