jNO: A JAX Library for Neural Operator and Foundation Model Training
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910207879151616 |
|---|---|
| author | Armbruster, Leon Ramesh, Rathan Kruse, Georg Straub, Christopher |
| author_facet | Armbruster, Leon Ramesh, Rathan Kruse, Georg Straub, Christopher |
| contents | jNO (jax Neural Operators) is a JAX-native library for neural operators and foundation models with unified support for both data-driven and physics-informed training. Its core design is a tracing system in which domains, model calls, residuals, supervised losses, and diagnostics are written in one symbolic language and compiled into one optimization pipeline. This allows users to move between operator regression, mesh-aware residual evaluation, and PDE-constrained training without restructuring the surrounding code. jNO also supports multi-model compositions, fine-grained control at parameter level (model, optimizer, and learning rate), hyperparameter tuning, and JAX-native workflows for translated PDE foundation-model families. The source repository is available at https://github.com/FhG-IISB/jNO. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_10159 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | jNO: A JAX Library for Neural Operator and Foundation Model Training Armbruster, Leon Ramesh, Rathan Kruse, Georg Straub, Christopher Machine Learning Numerical Analysis Computational Physics jNO (jax Neural Operators) is a JAX-native library for neural operators and foundation models with unified support for both data-driven and physics-informed training. Its core design is a tracing system in which domains, model calls, residuals, supervised losses, and diagnostics are written in one symbolic language and compiled into one optimization pipeline. This allows users to move between operator regression, mesh-aware residual evaluation, and PDE-constrained training without restructuring the surrounding code. jNO also supports multi-model compositions, fine-grained control at parameter level (model, optimizer, and learning rate), hyperparameter tuning, and JAX-native workflows for translated PDE foundation-model families. The source repository is available at https://github.com/FhG-IISB/jNO. |
| title | jNO: A JAX Library for Neural Operator and Foundation Model Training |
| topic | Machine Learning Numerical Analysis Computational Physics |
| url | https://arxiv.org/abs/2605.10159 |