⚡ Bolt: Optimize clustering algorithms with inline squared distance comparisons#133
⚡ Bolt: Optimize clustering algorithms with inline squared distance comparisons#133teerthsharma wants to merge 1 commit into
Conversation
Optimizes `auto_k_selection` and `region_query` in `aether-core::ml::clustering` by replacing exact distance computations (involving `sqrt`) with inline squared distance comparisons. Precomputes `epsilon * epsilon` and adds early exit conditions in the coordinate loop for significant performance gains during spatial scans. Also fixes a logic bug in `auto_k_selection` where `components` was incorrectly initialized and decremented instead of incremented. Co-authored-by: teerthsharma <78080953+teerthsharma@users.noreply.github.com>
|
👋 Jules, reporting for duty! I'm here to lend a hand with this pull request. When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down. I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job! For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with New to Jules? Learn more at jules.google/docs. For security, I will only act on instructions from the user who triggered this task. |
1 participant
💡 What: Replaced calls to
distance(which useslibm::sqrt) with inline squared distance comparisons (d^2 <= epsilon^2) inauto_k_selectionandregion_queryfunctions withincrates/aether-core/src/ml/clustering.rs. Fixed a state tracking bug inauto_k_selectionwherecomponentswas erroneously initialized tonand decremented, replacing it with an initialization to0and incrementing correctly. Added safeNaNhandling using negated comparison (!(sum < eps_sq)).🎯 Why: In density-based algorithms like DBSCAN (
region_query) and neighborhood graph builders (auto_k_selection), spatial distance checks are the primary bottleneck, exhibiting O(N^2) complexity behavior in dense regions. Computing the exact Euclidean distance with a square root is computationally expensive and unnecessary when only comparing against a constant threshold.📊 Impact: Expected to significantly reduce latency during clustering, especially in DBSCAN's region queries. It eliminates millions of
sqrtcalls and enables early loop termination as soon as the squared threshold is exceeded.🔬 Measurement: Verify by running
cargo test -p aether-core --offline ml::clustering::tests::test_auto_kand observing correct component counting. Measure latency of DBSCAN on large point clouds.PR created automatically by Jules for task 18212293129149971510 started by @teerthsharma