Kde loss dev - #126

Open

aphearin wants to merge 43 commits into

mainfrom

Commits on Jul 15, 2024

Add kdescent to testing env
aphearin
committed
ec88f83
Begin to build unit test of kdescent application
aphearin
committed
4cc5d72
flake8 fix
aphearin
committed
1fff82b
Add failing assert to check workflow setup
aphearin
committed
e7b2a72
Add test that should fail if kdescent is not installed
aphearin
committed
4d46271
Fix import issue
aphearin
committed
8575962
axopt --> optax
aphearin
committed
29ca8a3
Remove temporary assert
aphearin
committed
59fd622
Concatenate data
aphearin
committed
e3dd3a3
@alanpearl I have a working call to compare_kde_counts but the value of the loss is unexpectdly gigantic
aphearin
committed
8c1b015
Loss is gigantic even when fiducial and proposed models are extremely similar
aphearin
committed
ddc3025
Resolve my confusion over magnitude of kde-based loss
aphearin
committed
4924292
Restore large excursion in u_param space
aphearin
committed
76b4eb7

Commits on Jul 16, 2024

Commits on Jul 18, 2024

Commits on Jul 19, 2024

Add remaining kernels
aphearin
committed
6ff8c29

Commits on Jul 22, 2024

Commits on Jul 23, 2024

working call to single_sample_kde_loss_kern but results are singular
aphearin
committed
e653f9f
renormalize MAHs to zero to at lgm_obs
aphearin
committed
75fe5c1

Commits on Jul 24, 2024

Switch to fractional errors
aphearin
committed
0d07615
nan grad test passes
aphearin
committed
c152375
nan grad test passes for 100 different random u_params
aphearin
committed
22ef9ae
nan grad test passes for multi-objective loss
aphearin
committed
5b10f55
Add convenience function get_cens_target_data
aphearin
committed
fd658d3
Fix bug in test workflow
aphearin
committed
94d427d
Restore kdescent installation into workflow
aphearin
committed
74f2716
Add astropy into workflow
aphearin
committed
6c8e899
Remove hard-coding of NH_PER_M0BIN
aphearin
committed
1dcf2e0