⚡ Optimize Least-Squares Velocity Estimation - #7
Conversation
- Replaced manual normal equations solver with `np.linalg.lstsq` for better numerical robustness. - Improved efficiency by avoiding the creation of the large intermediate `gDotInv` matrix. - Preserved existing `sig_m` uncertainty estimation logic by calculating `gInv` specifically for the 2x2 design matrix. - Verified syntax and linting compliance. Co-authored-by: alfieprojectsdev <11991855+alfieprojectsdev@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. |
|
Closing: superseded by #22 which makes the same lstsq change to analysis.py but also retains gInv for uncertainty estimation. |
…P_DWLD from template Reconciliation debt: Option B (pre-download via igs_downloader, skip in-BPE FTP_DWLD) is half-built — template still ships 000 FTP_DWLD and campaign_builder.download_igs_products is defined but never called. Production PAGENET PCF already has no download step (matches Option B). RH-007 strips FTP_DWLD from the template, wires the pre-download into prepare_campaign(), and adds a pre-flight product-existence/naming check (gaps #4/#6/#7).
1 participant
💡 What: The optimization replaces the manual matrix inversion and explicit calculation of the
gDotInvmatrix withnp.linalg.lstsq.🎯 Why:
np.linalg.lstsq(typically using QR or SVD) is more robust against ill-conditioned matrices than solving the normal equations directly, which squares the condition number.(2, N)intermediate matrixgDotInv. For large datasets, this results in significant memory allocation and extra computation. By usinglstsq, we solve for the model parameters directly and avoid this allocation.📊 Measured Improvement:$O(N)$ vs $O(1)$ intermediate storage) and improved numerical robustness. Baseline measurement in the current environment was limited by missing dependencies, but the optimization follows standard numerical linear algebra best practices for least-squares problems.
The performance improvement is theoretically significant for large
Ndue to reduced memory overhead (PR created automatically by Jules for task 2650232839913107582 started by @alfieprojectsdev