Conversation
Keno
left a comment
There was a problem hiding this comment.
This is unsound as is, even if you want to make the assumption that taylor-ness is preserved. This just makes everything taylor which is unconditionally unsound.
|
in that case we definitely need test cases that break the assumption that everything is taylor because locally all the tests pass with this PR. (also, can you decipher the error message on CI I'm not able to understand or reproduce it). |
7da6082 to
9b56ede
Compare
Codecov ReportBase: 50.87% // Head: 50.77% // Decreases project coverage by
📣 This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more Additional details and impacted files@@ Coverage Diff @@
## main #109 +/- ##
==========================================
- Coverage 50.87% 50.77% -0.10%
==========================================
Files 27 27
Lines 2632 2572 -60
==========================================
- Hits 1339 1306 -33
+ Misses 1293 1266 -27
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
If we're willing to use the information that derivatives are linear operators, we can remove a lot of complexity from forward mode differentiation (and improve type stability in the process). I know @Keno doesn't believe this is valid to assume, but IMO the simplification enabled provides a pretty strong justification that this is the right approach.