search

HumanCompatibleAI / evaluating-rewards

Library to compare and evaluate reward functions
https://arxiv.org/abs/2006.13900
Apache License 2.0
61 stars 7 forks source link

Add tabular versions of new distance measures #19

Closed AdamGleave closed 4 years ago

AdamGleave commented 4 years ago

Adds support for computing new distance measures $D{asym}$ and $D{sym}$ in the tabular case, and integrates this into plot_gridworld_divergence.

If promising should add support in the function approximator setting as well. This may require more substantial modifications as it cannot just be derived from the loss, but rather requires loading the reward model.

codecov[bot] commented 4 years ago

Codecov Report

Merging #19 into master will not change coverage by %. The diff coverage is 100.00%.

Impacted file tree graph

@@            Coverage Diff            @@
##            master       #19   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            9        10    +1     
  Lines          371       463   +92     
=========================================
+ Hits           371       463   +92
Impacted Files Coverage Δ
tests/test_tabular.py 100.00% <100.00%> (ø)

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update d1a4ae1...3c8e7f4. Read the comment docs.