SciML / SciML/NeuralOperators.jl
Implement U-shaped Neural Operator (U-NO)
Nobody has claimed this yet.
- Dominant language
- Julia
- Stars
- 41
- Forks
- 15
- Avg merge
- 13h 14m
- Merged PRs (30d)
- 12
Description
Summary
Implement U-NO, which adapts the U-Net encoder-decoder architecture to neural operators in function spaces.
Reference
- Rahman et al., "U-NO: U-shaped Neural Operators," Transactions on Machine Learning Research, 2023. arXiv:2204.11127
Description
U-NO applies the U-Net multi-scale paradigm to neural operators, using integral operators that map between functions at different spatial scales. The encoder downsamples through spectral truncation, the decoder upsamples, and skip connections preserve high-frequency information. Reports up to 44% improvement on Navier-Stokes benchmarks over FNO.
This is architecturally related to the existing FNO implementation but requires multi-resolution spectral layers with downsampling/upsampling and skip connections between scales.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by locating the existing FNO implementation and its tests or entry points. Read the U-NO paper alongside the FNO code to define the multi-resolution spectral layers, downsampling and upsampling, and skip connections. Done means the U-NO architecture is implemented and its behavior is covered by appropriate tests, including the intended Navier–Stokes use case.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- julia
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100