Skip to content

Merge TCN, noise resiliency option, addition of more DIII-D shots, etc. from forked FRNN #48

@felker

Description

@felker

Mostly located on @ge-dong's forked version of the code.

  • Should the additional D3D shots be directly added to the existing list, or should it be kept separate for posterity with the Nature paper results? Kept separate, in both shot_lists/.txt files (d3d_clear_data_avail.txt vs.d3d_clear_since_2016.txt, e.g.) and internal variables in conf_parser.py (d3d_full vs. d3d_full_2019)
    • shot_lists/d3d_clear.txt and shot_lists/d3d_disrupt.txt are currently unused. Names of all files in shot_lists/ are confusing. See below for details.
  • Re-tune the hyperparameters for d3d_data and d3d_data_0D datasets on the expanded shot list.
(base) ➜  shot_lists wc -l d3d_clear*
  10000 d3d_clear_10000.txt
   1000 d3d_clear_1000.txt
    100 d3d_clear_100.txt
  18122 d3d_clear_data_avail.txt
  27311 d3d_clear_since_2016.txt
  74248 d3d_clear.txt
 130781 total
(base) ➜  shot_lists head -n1 d3d_clear*
==> d3d_clear_10000.txt <==
143800   -1.000000

==> d3d_clear_1000.txt <==
165500   -1.000000

==> d3d_clear_100.txt <==
167475   -1.000000

==> d3d_clear_data_avail.txt <==
125502   -1.000000

==> d3d_clear_since_2016.txt <==
125502   -1.000000

==> d3d_clear.txt <==
40560   -1.000000
(base) ➜  shot_lists tail -n1 d3d_clear*
==> d3d_clear_10000.txt <==
168554   -1.000000

==> d3d_clear_1000.txt <==
168554   -1.000000

==> d3d_clear_100.txt <==
168554   -1.000000

==> d3d_clear_data_avail.txt <==
168554   -1.000000

==> d3d_clear_since_2016.txt <==
180844   -1.000000

==> d3d_clear.txt <==
168554   -1.000000

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions