Cluster b2_* #44
Cluster b2_* #44
Conversation
There was a problem hiding this comment.
Pull request overview
This PR implements clustering analysis for force field parameters in response to issue #40. It introduces a new Python script that performs unsupervised clustering on bond and angle parameters from OFFXML force field files, identifying groups of similar parameters in (log(k), equilibrium) space.
Key Changes:
- New clustering.py script with comprehensive documentation and CLI interface using argparse
- Application to two force field variants:
b_no_bond_with_ring_in_atomandb2_no_bond_with_ring_in_atom - Output generation including CSV cluster assignments and PDF visualization plots
Reviewed changes
Copilot reviewed 7 out of 13 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| 8_clustering/clustering.py | New clustering script with HDBSCAN/DBSCAN algorithms for analyzing force field parameters |
| 8_clustering/b_no_bond_with_ring_in_atom/run.sh | Shell script to run clustering on b_no_bond_with_ring_in_atom force field |
| 8_clustering/b_no_bond_with_ring_in_atom/log.txt | Execution log showing 25 bond clusters and 7 angle clusters identified |
| 8_clustering/b_no_bond_with_ring_in_atom/clusters/bonds_clusters.csv | Bond parameter cluster assignments (554 parameters) |
| 8_clustering/b_no_bond_with_ring_in_atom/clusters/bonds_clusters.pdf | Visualization of bond parameter clusters |
| 8_clustering/b2_no_bond_with_ring_in_atom/run.sh | Shell script to run clustering on b2_no_bond_with_ring_in_atom force field |
| 8_clustering/b2_no_bond_with_ring_in_atom/log.txt | Execution log showing 35 bond clusters and 4 angle clusters identified |
| 8_clustering/b2_no_bond_with_ring_in_atom/clusters/bonds_clusters.pdf | Visualization of bond parameter clusters for b2 variant |
8_clustering/clustering.py
Outdated
| # Plot | ||
| fig, ax = plt.subplots(figsize=(8, 6)) | ||
| sorted_labels = list(sorted(set(labels))) | ||
| colors = plt.cm.get_cmap("nipy_spectral")( |
There was a problem hiding this comment.
Deprecated matplotlib API usage detected. The function plt.cm.get_cmap() was deprecated in Matplotlib 3.7 and will be removed in 3.11. Update to use matplotlib.colormaps[name] or matplotlib.colormaps.get_cmap() instead.
| # Angles tend to create concentric patterns, so use larger min_cluster_size by default | ||
| if method.lower() == "hdbscan": | ||
| if "min_cluster_size" not in kwargs_bonds: | ||
| kwargs_bonds["min_cluster_size"] = 3 |
There was a problem hiding this comment.
This expression mutates a default value.
| if "min_cluster_size" not in kwargs_bonds: | ||
| kwargs_bonds["min_cluster_size"] = 3 | ||
| if "min_cluster_size" not in kwargs_angles: | ||
| kwargs_angles["min_cluster_size"] = 3 |
There was a problem hiding this comment.
This expression mutates a default value.
| if "min_cluster_size" not in kwargs_angles: | ||
| kwargs_angles["min_cluster_size"] = 3 | ||
| if "min_samples" not in kwargs_angles: | ||
| kwargs_angles["min_samples"] = 5 # More conservative clustering |
There was a problem hiding this comment.
This expression mutates a default value.
This expression mutates a default value.
Close #40
Let me know if the copilot review was helpful or not and if there's something you wish it did.