- Most of the people listed for Action Items were not on the call
- Anarya's action item was not discussed in the East call. Anarya will attend East call if possible.
* Last week's East call - no agenda, talked about MRs.
* Quick updates (45 minutes)
- Operations (5 minutes)
- LL CBC operations: Made a new dashboard for Edward and Jacob analyses. Tried to add more info in one page. Fixed the high RAM we were seeing by adding patch related to reading the population model. Improved latency.
- Offline CBC operations: no update
- LL IDQ operations: RAM usage patch was a huge support for iDQ. Plots have improved a lot. Another patch to fix LR. Hope to get this in the next MDC. Will work on the two small analyses requested by review.
- O4 Dev (30 minutes)
- Low latency integrated testing and Monitoring: Becca is trying to add condor dag monitoring to Grafana and hook it up to alert.
- Template bank: no updates
- Likelihood ratio, background and foreground sampling: no updates
- Optimization and throughput benchmarking: no updates
- Data format wrangling: no updates
- Enabling running on OSG/IGWN grid: no updates
- DQ dev: no updates
- HM search: no updates
- Exploratory development (5 minutes): no updates
- Misc projects: no updates
- Paper Updates: no updates
* Focus session
- Manifold - Jim [slides](https://docs.google.com/presentation/d/1EIRJiiZq2yNQTjm2knIc-HRTAxFwBi_llIg1VBMRmTs/edit#slide=id.p)
- Questions
- Rachael: other applications?
- Interpolation between templates for RPE (led by Ryan)
- Ricci curvature, Riemann curvature requires more differentiations
- Rachael: is MANIFOLD template efficient for down stream calculations, like SVD?
- Unclear if SVD efficiency is gained by MANIFOLD placement
- MANIFOLD banks have nicer decompositions. Spectra is more monotonic than sbank.
- Jim, Ryan, and other are working on proving this.
- Becca: Has higher dimension parameter space been tried? Such as precession? Has other parameters shown to be more efficient?
- It is an arbitrary space. Python package is designed so you can give specific coordinates. It’s the same manifold, just different coordinates. The question is how is the diffeomorphism different between coordinates? Mu1 mu2 is being implemented now. Higher order modes and precession is work to be done.
- Becca: bank created with MANIFOLD what parameters?
- log(m1) * log(m2) * chi
- Prathamesh: does computational efficiency change with coordinate choice?
- Artifact of numerical features. Mismatch is sensitive to coordinate.
- If algorithm seems more curvature, it would place more coordinates.
- Shio: Why is the threshold chosen as 0.0001? (why did we settle on this particular value?) What was the motivation for implementing a minimum coordinate requirement?
- Motivation: corresponds to looking at the curvature of the space. If it’s a flat manifold, the distances are well controlled and you can sparesly sample the space and have mismatch threshold. In the opposite case, we need more templates to cover more curved space. It boiled down to volume in the coordinate.
- 0.0001: not sure
- Anarya: page 6: how are the boundaries of the cells chosen?
- Sbank temp placement is due to Voronoi placement.
- MANIFOLD: split up space so you don’t have to do all the match calculations.
