Store the slope in PointEvalGrad objects
#4
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…rue)`. Now, I noticed that the downstream `AbstractGaussNewton.terminate` method does not actually check the result when deciding whether to terminate. Rather than adjusting just the one solver, I decided to simply lift this condition into the top-level `_iterate.py`. As such we might consider whether we really need `.terminate` to return both `termintae: bool` and `result: RESULTS`. To consider what each combination means: - `terminate=True`, `result=RESULTS.successful`: done, and all went well. - `terminate=True`, `result=RESULTS.foo`: something went wrong; halt. - `terminate=False`, `result=RESULTS.successfull`: keep making more iterations. - `terminate=False`, `result=RESULTS.foo`: <invalid> So we really do need two return arguments: it's not enough to just stop if `result != RESULTS.successful`, because otherwise we have no way to communicate success-and-terminate. Now we might wonder whether this change produces a change in behaviour. I think technically yes: a solver may have set `terminate=False` and `result=RESULTS.foo`, and then later decided to switch back to being successful. In practice I don't think we ever did that, and it seems pretty sketchy anyway.
Fixed Gauss-Newton including a spurious `lx.linear_solve(..., throw=True)`
…_fix Support Optax solvers that include a linesearch
… using Zoom In the accepted branch avoid recalculating the gradient, and reuse it reading from ZoomState.current_point.grad as this is the gradient at y_eval if y_eval was accepted.
Suggested by all sources and required for test_compat to pass with BFGS.
Instead of passing lin_fn and options to searches, add a _needs_grad_at_y_eval to the searches. In AbstractQuasiNewton and AbstractGradientDescent: if the search needs the gradient, calculate it and include it in f_eval_info.
johannahaffner
force-pushed
the
zoom_linesearch
branch
from
2cfa5d4 to
00e82f4
Compare
johannahaffner
force-pushed
the
zoom_linesearch
branch
2 times, most recently
from
dd40198 to
032782c
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.