Topology suppression

[augustocattafesta]

No description provided.

[Copilot]

Pull request overview

Adds “topology suppression” to the clustering pipeline to zero-out pixels that don’t match an expected neighborhood pattern, aiming to improve cluster quality before reconstruction.

Changes:

• Introduces ClusteringBase.topology_suppress() to suppress pixels that don’t match the expected neighbor topology and re-sort outputs.
• Integrates topology suppression into ClusteringNN.run() prior to trimming zero-suppressed pixels.

Comments suppressed due to low confidence (1)

src/hexsample/clustering.py:331

• After topology_suppress, pha, col, and row have already been sliced/reordered to match the selected pixels. Reusing mask (which still contains indices into the original 7-pixel arrays) to index col/row/pha will produce incorrect selections and can raise IndexError when mask contains values > len(col)-1. Use a boolean keep mask derived from the post-suppression arrays (e.g., keep = pha > 0) and apply it directly to pha/col/row, or reset mask to np.arange(len(pha))[pha > 0] before indexing.

        mask = mask[pha > 0]
        # Trim the relevant arrays.
        col = col[mask]
        row = row[mask]
        pha = pha[mask]

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.