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

[andorsk]

Copied from docs:

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]

Early results:

Black Hole - Black Hole Intra-class Comparison
Average similarity Untuned + No Domain Expertise : 0.9612167141597305
Average similarity Tuned + Domain Expertise: 0.9264347568287536
Tuning done on all data:

[andorsk]

Migrating issue to animikhroy/rk_toolkit_pipeline_diagrams#11