Add Pantara v7d — physics law discovery system - #211
Open
Yapock22 wants to merge 5 commits into
Open
Conversation
Pantara is a matching-pursuit symbolic regression system for physical laws combining a neural oracle with analytical regression. Key properties: - Analytical power-law detection in O(N) via log-log least-squares - 8 function families × 6 transformation spaces - Pre-trained SetEncoder oracle (15k synthetic episodes) - Scikit-learn fit/predict interface; eval_kwargs disables scaling so that positive-domain log-log detection works correctly SRBench results (code frozen, no post-hoc tuning): 132/133 Feynman+Strogatz datasets evaluated R²≥0.90 on 86/132 datasets (65%) R²=1.000 on 37 datasets (pure power laws, <0.2 s each) Mean R²=0.641, median R²=0.970 Source: https://github.com/Yapock22/pantara Install: pip install git+https://github.com/Yapock22/pantara.git Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Member
|
thanks for your patience. we finally fixed the other methods that were failing in docker on main. once the tests pass we'll perform our review and update you. |
lacava
added a commit
that referenced
this pull request
Aug 13, 2026
CONTRIBUTING.md and the PR template still described the old single-directory layout, from before regressor.py moved into experiment/methods/ (b5c5c96). back then CI copied the files over via scripts/copy_algorithm_files.sh, so contributors only touched algorithms/. that script is gone but the docs never got updated, and nothing catches the mistake: the build-and-test matrix comes from `ls algorithms/`, so a method added only under experiment/methods/ never gets a job and still goes all green (see #211). - CONTRIBUTING.md documents where files go and why it's split that way - PR template reorganized around the two directories - local_ci.sh now wraps the docker commands instead of the old conda flow - scripts/check_method_layout.py checks the two dirs agree; runs in CI as validate-layout and gates build-and-test - fixes algorithms/eql/metadata.yml (invalid yaml), a missing experiment/methods/xgboost/__init__.py, and an invalid escape sequence in gplearn's regressor.py
lacava
added a commit
that referenced
this pull request
Aug 13, 2026
CONTRIBUTING.md and the PR template still described the old single-directory layout, from before regressor.py moved into experiment/methods/ (b5c5c96). back then CI copied the files over via scripts/copy_algorithm_files.sh, so contributors only touched algorithms/. that script is gone but the docs never got updated, and nothing catches the mistake: the build-and-test matrix comes from `ls algorithms/`, so a method added only under experiment/methods/ never gets a job and still goes all green (see #211). - CONTRIBUTING.md documents where files go and why it's split that way - PR template reorganized around the two directories - local_ci.sh now wraps the docker commands instead of the old conda flow - scripts/check_method_layout.py checks the two dirs agree; runs in CI as validate-layout and gates build-and-test - fixes algorithms/eql/metadata.yml (invalid yaml), a missing experiment/methods/xgboost/__init__.py, and an invalid escape sequence in gplearn's regressor.py
lacava
added a commit
that referenced
this pull request
Aug 13, 2026
CONTRIBUTING.md and the PR template still described the old single-directory layout, from before regressor.py moved into experiment/methods/ (b5c5c96). back then CI copied the files over via scripts/copy_algorithm_files.sh, so contributors only touched algorithms/. that script is gone but the docs never got updated, and nothing catches the mistake: the build-and-test matrix comes from `ls algorithms/`, so a method added only under experiment/methods/ never gets a job and still goes all green (see #211). - CONTRIBUTING.md documents where files go and why it's split that way - PR template reorganized around the two directories - local_ci.sh now wraps the docker commands instead of the old conda flow - scripts/check_method_layout.py checks the two dirs agree; runs in CI as validate-layout and gates build-and-test - fixes algorithms/eql/metadata.yml (invalid yaml), a missing experiment/methods/xgboost/__init__.py, and an invalid escape sequence in gplearn's regressor.py
Member
|
The CI failed because there is experiment/methods/pantara/ (the method wrapper) but no corresponding algorithms/pantara/ folder. The CI matrix is built from the algorithms/ directory, so any method without algorithms// is never built or tested — the validator test now flags this as a layout problem. see updated https://cavalab.org/srbench/contributing/ guide. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Pantara is a physics law discovery system that identifies
mathematical structure from raw sensor data using statistical
signature detection, analytical power-law regression (lstsq
in log space), and a specialized trigonometric estimator.
Evaluated on 132 Feynman/Strogatz datasets from SRBench:
Best on multiplicative/power-law equations. Does not cover
Gaussian distributions, Euclidean distances, or differential
equations.
Code: github.com/Yapock22/pantara