Updated intra-class similarity scores between BH-BH binaries with Loss Optimization Graph #6

[andorsk]

Linked Issue: https://github.com/andorsk/rk_toolkit/issues/6

[animikhroy]

I have made a doc to collect all the results, images and brief descriptions in one place with relevant comments wherever applicable for your consideration when you address this issue today. Here is the link to that results doc for your reference:

https://docs.google.com/document/d/13DjS3Y44cE9hJq6H403zvSlkI1o9uMNBhBhK9jaRGQ8/edit?usp=sharing

This google doc is solely focused on the updated results and descriptions for ligo_secondary_analysis.ipynb. I am putting all the issues in sequence and requesting brief descriptions of techniques and methodologies involved with relevant references to cite (if any) based on your work, for me to then elaborate further on those points in details and quickly wrap up the final draft of the paper for peer review.

[andorsk]

Resending notes here and migrating issue:

You had sent me some preliminary pre-tuned and post tuned similarity measures for intra-class BH-BH similarity scores as follows:

Pre-turned Similarity Measure: Average similarity of graphs pre tuned: 0.9565655957632736

Post-tuned SImilarity Measure: Iteration: 01800. Loss: 0.89918207. ITime 20.91 seconds. Total time: 371.77

I assume this is where you had worked on it as shared with me on whatsapp: https://ml.kesselmanrao.com/notebooks/ligo_v2/livgo_v2.ipynb But I can't find the graph and results there anymore so I assume you wanted to redo things properly and started working on it after I left for the UK as evident in the sections of the notebook. So I would request you to revisit the notebook and send me the updated intra-class BH-BH scores and loss optimization graph for inclusion in the paper.

[andorsk]

2000 iterations over 2 min. Final loss: 0.84397022 <- across all.

Iteration: 00100. Loss: 0.94509804. ITime 6.24 seconds. Total time: 6.36
Iteration: 00200. Loss: 0.91825397. ITime 6.48 seconds. Total time: 12.84
Iteration: 00300. Loss: 0.90267000. ITime 6.61 seconds. Total time: 19.45
Iteration: 00400. Loss: 0.89582808. ITime 6.50 seconds. Total time: 25.95
Iteration: 00500. Loss: 0.87988594. ITime 6.55 seconds. Total time: 32.50
Iteration: 00600. Loss: 0.92906266. ITime 6.65 seconds. Total time: 39.15
Iteration: 00700. Loss: 0.88783736. ITime 6.67 seconds. Total time: 45.82
Iteration: 00800. Loss: 0.92518613. ITime 6.85 seconds. Total time: 52.67
Iteration: 00900. Loss: 0.92272765. ITime 6.55 seconds. Total time: 59.22
Iteration: 01000. Loss: 0.88411593. ITime 6.72 seconds. Total time: 65.94
Iteration: 01100. Loss: 0.90092668. ITime 6.69 seconds. Total time: 72.62
Iteration: 01200. Loss: 0.89094607. ITime 6.61 seconds. Total time: 79.23
Iteration: 01300. Loss: 0.88140951. ITime 6.59 seconds. Total time: 85.83
Iteration: 01400. Loss: 0.88584417. ITime 6.74 seconds. Total time: 92.57
Iteration: 01500. Loss: 0.88708139. ITime 6.69 seconds. Total time: 99.26
Iteration: 01600. Loss: 0.86845898. ITime 6.54 seconds. Total time: 105.80
Iteration: 01700. Loss: 0.87034554. ITime 6.68 seconds. Total time: 112.48
Iteration: 01800. Loss: 0.90068736. ITime 6.62 seconds. Total time: 119.10
Iteration: 01900. Loss: 0.84397022. ITime 6.61 seconds. Total time: 125.70

Training occurs across all rk models.

BH-BH:

Average similarity Untuned: 0.9007671115626494 Average similarity Tuned: 0.8767771992341131

[andorsk]

@animikhroy answered on the sheet. Take a look and lmk if I can close the issue. I would like to do a revisit if I have time, but I think moving on to the other items is more important right now.

[animikhroy]

@andorsk I have read your comments on the google doc on this and I would suggest 2 things:

1) there was a NS-NS similarity test score you had generated earlier. Could you paste it there as well? Then I can use both as intra class similarity scores for BH-BH as well as NS-NS.

2) Let's revisit this in the end if there is more time. Order of priority wise I would say combining and compiling the ligo_primary_analysis.ipynb notebook is more important use of your limited time, after that we can consider revisiting this result.

Thanks!

[animikhroy]

This issue was closed upon internal review. Will reopen based on any specific requirements from the peer-reviewers.