Update PDHG docstring.#2220
Open
manchester-jhellier wants to merge 2 commits intoTomographicImaging:masterfrom
Open
Update PDHG docstring.#2220manchester-jhellier wants to merge 2 commits intoTomographicImaging:masterfrom
manchester-jhellier wants to merge 2 commits intoTomographicImaging:masterfrom
Conversation
MargaretDuff
reviewed
Oct 3, 2025
Member
MargaretDuff
left a comment
MargaretDuff
left a comment
There was a problem hiding this comment.
Thanks @manchester-jhellier for starting this conversation! I made some comments
|
|
||
| The general problem considered in the PDHG algorithm is the generic saddle-point problem | ||
|
|
||
| .. math:: \min_{x\in X}\max_{y\in Y} \langle Kx, y \rangle + g(x) - f^{*}(x) |
Member
There was a problem hiding this comment.
Your above comment led me to spot this...
Suggested change
| .. math:: \min_{x\in X}\max_{y\in Y} \langle Kx, y \rangle + g(x) - f^{*}(y) |
| A Linear Operator. | ||
| sigma : positive :obj:`float`, or `np.ndarray`, `DataContainer`, `BlockDataContainer`, optional, default is 1.0/norm(K) or 1.0/ (tau*norm(K)**2) if tau is provided | ||
| Step size for the dual problem. | ||
| Step size for the dual problem. Needs to obey constraints with tau and operator norm to be valid, see below for details. |
Member
There was a problem hiding this comment.
Suggested change
| Step size for the dual problem. Needs to obey constraints with tau and operator norm to be valid, see below for details. | |
| Step size for the dual problem. Needs to obey constraints with tau and operator norm to satisfy convergence guarantees, see below for details. |
Potentially? Valid is interesting as you get step sizes that don't meet the convergence guarantees but still lead to convergence... and this can vastly speed things up!
| Step size for the dual problem. Needs to obey constraints with tau and operator norm to be valid, see below for details. | ||
| tau : positive :obj:`float`, or `np.ndarray`, `DataContainer`, `BlockDataContainer`, optional, default is 1.0/norm(K) or 1.0/ (sigma*norm(K)**2) if sigma is provided | ||
| Step size for the primal problem. | ||
| Step size for the primal problem. Needs to obey constraints with sigma and operator's norm to be valid, see below for details. |
Member
There was a problem hiding this comment.
Suggested change
| Step size for the primal problem. Needs to obey constraints with sigma and operator's norm to be valid, see below for details. | |
| Step size for the primal problem. Needs to obey constraints with sigma and operator's norm to satisfy convergence guarantees, see below for details. |
| ---------- | ||
| f : Function | ||
| A convex function with a "simple" proximal method of its conjugate. | ||
| A convex function with a "simple" proximal method of its conjugate. This function must map from the operator range to the Reals, as :math: `f(Kx)` or :math: `f^{*}(x)` will be the contribution to the total objective. See below for details. |
Member
There was a problem hiding this comment.
Suggested change
| A convex function with a "simple" proximal method of its conjugate. This function must map from the operator range to the Reals, as :math: `f(Kx)` or :math: `f^{*}(x)` will be the contribution to the total objective. See below for details. | |
| A convex function with a "simple" proximal method of its conjugate. This function must map from the operator range to the Reals, as :math: `f(Kx)` will be the contribution to the total objective. See below for details. |
I think leave the convex conjugate out for now?
|
|
||
| class PDHG(Algorithm): | ||
|
|
||
| r"""Primal Dual Hybrid Gradient (PDHG) algorithm, see :cite:`CP2011`, :cite:`EZXC2010`. |
Member
There was a problem hiding this comment.
Maybe if we make the objective clear at the top it will also help?
Suggested change
| r"""Primal Dual Hybrid Gradient (PDHG) algorithm, see :cite:`CP2011`, :cite:`EZXC2010`. | |
| PDHG minimises objectives of the form: | |
| .. math:: \min_{x\in X} f(Kx) + g(x), | |
| where :math:`f` and the regulariser :math:`g` need to be proper, convex and lower semi-continuous. The function :math:`f` and the convex conjugate of :math:`g` must also have calculable proximal methods. |
Member
|
Hi @manchester-jhellier, thanks for opening this PR. Are you happy for us to commit Margaret's suggested changes? 😄 |
gfardell
added
the
community-contribution
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
Description
Updated the docstring for PDHG to hopefully make it a bit clearer. Feel free to change. Related issue is #2219 .
Contribution Notes
❤️ Thanks for your contribution!