search

jacopok / mlgw_bns

Accelerating gravitational wave template generation with machine learning.
GNU General Public License v3.0
4 stars 2 forks source link

Test multibanding #5

Closed jacopok closed 2 years ago

jacopok commented 2 years ago

When using multibanding it seems that sometimes the reconstruction mismatches are very high; is this a "real" problem or an implementation bug?

Also, we need more tests for it.

jacopok commented 2 years ago

The issue might be connected to the non-power-of-2 nature of the df computed by mlgw_bns.multibanding.seglen_from_freq.

Fixing that, and unifying it with mlgw_bns.dataset_generation.Dataset.optimal_df_hz is the priority.

jacopok commented 2 years ago

Also, the EOB waveforms I'm comparing to are not computed properly, they are uniform in frequency!

jacopok commented 2 years ago

The aforementioned issues are all solved by https://github.com/jacopok/mlgw_bns/commit/852591d6c9597916b19e32cefe3fd27148579c0e, but I will not close this yet since I need to test multibanding with the reconstruction of sub-20Hz waveforms.

jacopok commented 2 years ago

This is now done, since the default dataset is multibanded since https://github.com/jacopok/mlgw_bns/commit/3ff3cc09666a5e994d8bc68a49d37a5014631751