Data and Python code for angular collision diversity in three-dimensional point clouds
Description
Python code and numerical results for finite-cloud angular collision diversity at a stated resolution. The deposit includes six-family simulations, bandwidth sensitivity, exact-target calibration, sufficient stability-bound checks, comparisons with binned angular collision and a scalar distance kernel, and public illustrations using Stanford Bunny, Armadillo and Drill. The angular kernel has smaller median perturbation changes in the specified comparison, while the distance kernel has higher held-out classification accuracy. Public-object analyses distinguish full-vertex sampling, uniform subsets, noisy subsets and density-biased subsets. Original Stanford models are not redistributed; verified downloaders and source checksums identify the inputs. Repeated numerical runs are not independent physical objects. Conditional Monte Carlo uncertainty does not establish population-sampling coverage.
Files
Steps to reproduce
1. Download and extract Mendeley_Data_P1.zip to a writable directory. Open a terminal in the extracted directory containing requirements.txt and README.md. 2. Create a clean Python environment compatible with ENVIRONMENT.txt. Install dependencies with: python -m pip install -r requirements.txt 3. Preserve a copy of the frozen outputs before running any generator. The master seed is 34071985; task-specific offsets and parameters are explicit in the source. 4. Run implementation checks: python revision_experiments.py --task tests 5. Obtain the external model from its original source: python fetch_bunny.py This verifies the downloaded archive and extracts the recorded Stanford Bunny member. Read THIRD_PARTY_NOTICE.txt. Internet access is needed for this download; the source model is not part of this deposit. 6. Run the additional experiments: python revision_experiments.py --task p1 This generates the bandwidth, exact-target calibration and public-object summaries in new_results/ and figures in new_figures/. Without the downloaded model, the public example cannot be reproduced. 7. To regenerate the original reference experiment for this study only, run: python -c "import sys; sys.path.insert(0,'core'); import run_studies as r; r.setup(); r.paper1()" Its outputs are written to core/results/, core/figures/ and core/tables/. Compare them with the corresponding frozen files in legacy_results/. Do not run the full four-study pipeline merely to reproduce one study. 8. Compare generated numerical summaries with the frozen CSV files and verify file integrity using SHA256SUMS.txt. Timing is hardware dependent; last-bit numerical differences and evolutionary trajectories may differ across platforms. Separate revised outputs from legacy outputs; they are not interchangeable. ADDITIONAL EXPANDED ANALYSES After installing requirements, run python fetch_extension_inputs.py --output external, then python extension_analysis.py --inputs external. This recreates extension_results with seed 34081985. The downloads retain Stanford terms. Consult extension_results/design.json for budgets and replicate counts.