Skip to content

Use exact solver for cuML t-SNE to avoid Barnes-Hut collapse #45

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
NetZissou merged 1 commit into from
Jun 29, 2026
Merged

Use exact solver for cuML t-SNE to avoid Barnes-Hut collapse #45

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 5 additions & 2 deletions shared/utils/clustering.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -250,10 +250,13 @@ def _reduce_dim_cuml(embeddings: np.ndarray, method: str, seed: Optional[int], n
n_samples = embeddings.shape[0]
perplexity = min(30, max(5, n_samples // 3))

# Force the exact solver: cuML's default Barnes-Hut collapses to a
# ~1D line on near-homogeneous data (#40). exact is O(N^2) but fine
# at our interactive scale; a faster Barnes-Hut-with-guard can come later.
if seed is not None:
reducer = cuTSNE(n_components=2, perplexity=perplexity, random_state=seed)
reducer = cuTSNE(n_components=2, perplexity=perplexity, method="exact", random_state=seed)
else:
reducer = cuTSNE(n_components=2, perplexity=perplexity)
reducer = cuTSNE(n_components=2, perplexity=perplexity, method="exact")
else:
raise ValueError("Unsupported method. Choose 'PCA', 'TSNE', or 'UMAP'.")

Expand Down
Loading