godofecht / godofecht/flow-scikit
[Perf #478] KernelRidge RBF predict: eliminate ~7.7x inference gap with batched kernel evaluation
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 2
- Forks
- 0
- Avg merge
- 4h 33m
- Merged PRs (30d)
- 9
Description
Parent: #478
Canonical v2 Diabetes: Flow predict 4.861 ms vs sklearn 0.628 ms. Unlike several other losing estimators, both training and inference are independently slow, so prediction needs its own fix.
Profile test-to-train RBF kernel construction, squared-distance computation, exponentials, alpha dot products, batching, vectorization and temporary matrices. Audit whether Flow materializes more data than required and whether the final kernel-vector product can use optimized dense kernels.
Acceptance: prediction-only profile; batch/sample/support-size curves; zero avoidable per-element allocation; parity remains verified; canonical predict >=0.95x first, target >=1.05x.
Contributor guide
No contributing guide indexed for this repository
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
Locate the KernelRidge RBF predict implementation and begin with a prediction-only profile on the canonical v2 Diabetes case. Measure kernel construction, squared distances, exponentials, alpha products, batching, vectorization, and temporary matrices across the requested curves; done means parity remains verified and the canonical prediction reaches at least 0.95x, with 1.05x as the target.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, scikit-learn
- Domain
- machine-learning, performance
- Issue type
- Refactor
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100