Neptuna framework benchmarks complex multiphase flows with large dataset

Researchers have developed Neptuna, a new machine learning framework designed to benchmark complex multiphase flows, which are crucial in applications like bubble collapse and droplet breakup. The framework includes a substantial dataset of 2.4 TB of high-fidelity 2D and 3D data. Various model families, including transformers and pre-trained PDE foundation models, were evaluated using standard and composite loss functions, with SoftAdapt showing consistent improvements in interface preservation and spectral fidelity. AI

IMPACT Introduces a new benchmark and dataset for evaluating ML models on complex fluid dynamics, potentially advancing AI applications in physics simulations.

RANK_REASON The item describes a new research paper introducing a machine learning framework and dataset for benchmarking complex multiphase flows. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• GradNorm
• Harish Ramachandran
• Hugging Face
• ScienceCast
• SoftAdapt

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →