Rework LevenbergMarquardt to use the vector function functionality

Changelog.md

@@ -7,9 +7,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [0.5.5] unreleased
 
+### Added
+
+* the Levenberg-Marquardt algorithm internally uses a `VectorGradientFunction`, which allows
+ to use a vector of gradients of a function returning all gradients as wel for the algorithm
+* The `VectorGradientFunction` now also have a `get_jacobian` function
+
 ### Changed
 
 * Minimum Julia version is now 1.10 (the LTS which replaced 1.6)
+* The vectorial functions had a bug where the original vector function for the mutating case
+  was not always treated as mutating.
 
 ### Removed
 
@@ -20,7 +28,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Added
 
 * An automated detection whether the tutorials are present
-   if not an also no quarto run is done, an automated `--exlcude-tutorials` option is added.
+   if not an also no quarto run is done, an automated `--exclude-tutorials` option is added.
 * Support for ManifoldDiff 0.4
 * icons upfront external links when they link to another package or wikipedia.
 
@@ -32,7 +40,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Changed
 
-* stabilize `max_Stepzise` to also work when `injectivity_radius` dos not exist.
+* stabilize `max_stepsize` to also work when `injectivity_radius` dos not exist.
   It however would warn new users, that activate tutorial mode.
 * Start a `ManoptTestSuite` subpackage to store dummy types and common test helpers in.
 
@@ -71,8 +79,8 @@ In general we introduce a few factories, that avoid having to pass the manifold
 
 ### Changed
 
-* Any `Stepsize` now hase a `Stepsize` struct used internally as the original `struct`s before. The newly exported terms aim to fit `stepsize=...` in naming and create a `ManifoldDefaultsFactory` instead, so that any stepsize can be created without explicitly specifying the manifold.
-  * `ConstantStepsize` is no longer exported, use `ConstantLength` instead. The length parameter is now a positional argument following the (optonal) manifold. Besides that `ConstantLength` works as before,just that omitting the manifold fills the one specified in the solver now.
+* Any `Stepsize` now has a `Stepsize` struct used internally as the original `struct`s before. The newly exported terms aim to fit `stepsize=...` in naming and create a `ManifoldDefaultsFactory` instead, so that any stepsize can be created without explicitly specifying the manifold.
+  * `ConstantStepsize` is no longer exported, use `ConstantLength` instead. The length parameter is now a positional argument following the (optional) manifold. Besides that `ConstantLength` works as before,just that omitting the manifold fills the one specified in the solver now.
   * `DecreasingStepsize` is no longer exported, use `DecreasingLength` instead. `ConstantLength` works as before,just that omitting the manifold fills the one specified in the solver now.
   * `ArmijoLinesearch` is now called `ArmijoLinesearchStepsize`. `ArmijoLinesearch` works as before,just that omitting the manifold fills the one specified in the solver now.
   * `WolfePowellLinesearch` is now called `WolfePowellLinesearchStepsize`, its constant `c_1` is now unified with Armijo and called `sufficient_decrease`, `c_2` was renamed to `sufficient_curvature`. Besides that, `WolfePowellLinesearch` works as before, just that omitting the manifold fills the one specified in the solver now.
@@ -134,7 +142,7 @@ In general we introduce a few factories, that avoid having to pass the manifold
     * `ExactPenaltyMethodState(M, sub_problem; evaluation=...)` was added and `ExactPenaltyMethodState(M, sub_problem, sub_state; evaluation=...)` now has `p=rand(M)` as keyword argument instead of being the second positional one
     * `DifferenceOfConvexState(M, sub_problem; evaluation=...)` was added and `DifferenceOfConvexState(M, sub_problem, sub_state; evaluation=...)` now has `p=rand(M)` as keyword argument instead of being the second positional one
     * `DifferenceOfConvexProximalState(M, sub_problem; evaluation=...)` was added and `DifferenceOfConvexProximalState(M, sub_problem, sub_state; evaluation=...)` now has `p=rand(M)` as keyword argument instead of being the second positional one
-  * bumped `Manifolds.jl`to version 0.10; this mainly means that any algorithm working on a productmanifold and requiring `ArrayPartition` now has to explicitly do `using RecursiveArrayTools`.
+  * bumped `Manifolds.jl`to version 0.10; this mainly means that any algorithm working on a product manifold and requiring `ArrayPartition` now has to explicitly do `using RecursiveArrayTools`.
 ### Fixed
 
 * the `AverageGradientRule` filled its internal vector of gradients wrongly – or mixed it up in parallel transport. This is now fixed.

docs/src/plans/objective.md

@@ -121,6 +121,8 @@ ManifoldCostGradientObjective
 ```@docs
 get_gradient
 get_gradients
+get_residuals
+get_residuals!
 ```
 
 and internally
@@ -129,12 +131,6 @@ and internally
 get_gradient_function
 ```
 
-#### Internal helpers
-
-```@docs
-get_gradient_from_Jacobian!
-```
-
 ### Subgradient objective
 
 ```@docs
@@ -262,6 +258,8 @@ Manopt.FunctionVectorialType
 #### Access functions
 
 ```@docs
+Manopt.get_jacobian
+Manopt.get_jacobian!
 Manopt.get_value
 Manopt.get_value_function
 Base.length(::VectorGradientFunction)
@@ -271,6 +269,9 @@ Base.length(::VectorGradientFunction)
 
 ```@docs
 Manopt._to_iterable_indices
+Manopt._change_basis!
+Manopt.get_basis
+Manopt.get_range
 ```
 
 ### Subproblem objective

docs/src/solvers/ChambollePock.md

@@ -116,7 +116,7 @@ The [`ChambollePock`](@ref) solver requires the following functions of a manifol
 * A [`retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, q, p, X)`; it is recommended to set the [`default_retraction_method`](@extref `ManifoldsBase.default_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `retraction_method=` or `retraction_method_dual=` (for ``\mathcal N``) does not have to be specified.
 * An [`inverse_retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, X, p, q)`; it is recommended to set the [`default_inverse_retraction_method`](@extref `ManifoldsBase.default_inverse_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `inverse_retraction_method=` or `inverse_retraction_method_dual=` (for ``\mathcal N``) does not have to be specified.
 * A [`vector_transport_to!`](@extref ManifoldsBase :doc:`vector_transports`)`M, Y, p, X, q)`; it is recommended to set the [`default_vector_transport_method`](@extref `ManifoldsBase.default_vector_transport_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `vector_transport_method=` or `vector_transport_method_dual=` (for ``\mathcal N``) does not have to be specified.
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 
 ## Literature
 

docs/src/solvers/DouglasRachford.md

@@ -74,7 +74,7 @@ The [`DouglasRachford`](@ref) solver requires the following functions of a manif
 
 * A [`retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, q, p, X)`; it is recommended to set the [`default_retraction_method`](@extref `ManifoldsBase.default_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `retraction_method=` does not have to be specified.
 * An [`inverse_retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, X, p, q)`; it is recommended to set the [`default_inverse_retraction_method`](@extref `ManifoldsBase.default_inverse_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `inverse_retraction_method=` does not have to be specified.
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 
 By default, one of the stopping criteria is [`StopWhenChangeLess`](@ref),
 which requires

docs/src/solvers/LevenbergMarquardt.md

@@ -21,7 +21,7 @@ The [`LevenbergMarquardt`](@ref) solver requires the following functions of a ma
 
 * A [`retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, q, p, X)`; it is recommended to set the [`default_retraction_method`](@extref `ManifoldsBase.default_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `retraction_method=` does not have to be specified.
 * the [`norm`](@extref `LinearAlgebra.norm-Tuple{AbstractManifold, Any, Any}`) as well, to stop when the norm of the gradient is small, but if you implemented `inner`, the norm is provided already.
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 
 
 ## Literature

docs/src/solvers/augmented_Lagrangian_method.md

@@ -26,7 +26,7 @@ AugmentedLagrangianGrad
 
 The [`augmented_Lagrangian_method`](@ref) solver requires the following functions of a manifold to be available
 
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 * Everything the subsolver requires, which by default is the [`quasi_Newton`](@ref) method
 * A [`zero_vector`](@extref `ManifoldsBase.zero_vector-Tuple{AbstractManifold, Any}`)`(M,p)`.
 

docs/src/solvers/difference_of_convex.md

@@ -67,7 +67,7 @@ which either requires
 
 * A [`retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, q, p, X)`; it is recommended to set the [`default_retraction_method`](@extref `ManifoldsBase.default_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `retraction_method=` or `retraction_method_dual=` (for ``\mathcal N``) does not have to be specified.
 * An [`inverse_retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, X, p, q)`; it is recommended to set the [`default_inverse_retraction_method`](@extref `ManifoldsBase.default_inverse_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `inverse_retraction_method=` or `inverse_retraction_method_dual=` (for ``\mathcal N``) does not have to be specified or the [`distance`](@extref `ManifoldsBase.distance-Tuple{AbstractManifold, Any, Any}`)`(M, p, q)` for said default inverse retraction.
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 * By default the tangent vector storing the gradient is initialized calling [`zero_vector`](@extref `ManifoldsBase.zero_vector-Tuple{AbstractManifold, Any}`)`(M,p)`.
 * everything the subsolver requires, which by default is the [`trust_regions`](@ref) or if you do not provide a Hessian [`gradient_descent`](@ref).
 

docs/src/solvers/exact_penalty_method.md

@@ -30,7 +30,7 @@ LogarithmicSumOfExponentials
 The [`exact_penalty_method`](@ref) solver requires the following functions of a manifold to be available
 
 
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 * Everything the subsolver requires, which by default is the [`quasi_Newton`](@ref) method
 * A [`zero_vector`](@extref `ManifoldsBase.zero_vector-Tuple{AbstractManifold, Any}`)`(M,p)`.
 

docs/src/solvers/particle_swarm.md

@@ -30,7 +30,7 @@ The [`particle_swarm`](@ref) solver requires the following functions of a manifo
 * A [`vector_transport_to!`](@extref ManifoldsBase :doc:`vector_transports`)`M, Y, p, X, q)`; it is recommended to set the [`default_vector_transport_method`](@extref `ManifoldsBase.default_vector_transport_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `vector_transport_method=` does not have to be specified.
 * By default the stopping criterion uses the [`norm`](@extref `LinearAlgebra.norm-Tuple{AbstractManifold, Any, Any}`) as well, to stop when the norm of the gradient is small, but if you implemented `inner`, the norm is provided already.
 * Tangent vectors storing the social and cognitive vectors are initialized calling [`zero_vector`](@extref `ManifoldsBase.zero_vector-Tuple{AbstractManifold, Any}`)`(M,p)`.
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 * The [`distance`](@extref `ManifoldsBase.distance-Tuple{AbstractManifold, Any, Any}`)`(M, p, q)` when using the default stopping criterion, which uses [`StopWhenChangeLess`](@ref).
 
 ## Literature

docs/src/solvers/primal_dual_semismooth_Newton.md

@@ -78,7 +78,7 @@ The [`primal_dual_semismooth_Newton`](@ref) solver requires the following functi
 * A [`retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, q, p, X)`; it is recommended to set the [`default_retraction_method`](@extref `ManifoldsBase.default_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `retraction_method=` does not have to be specified.
 * An [`inverse_retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, X, p, q)`; it is recommended to set the [`default_inverse_retraction_method`](@extref `ManifoldsBase.default_inverse_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `inverse_retraction_method=` does not have to be specified.
 * A [`vector_transport_to!`](@extref ManifoldsBase :doc:`vector_transports`)`M, Y, p, X, q)`; it is recommended to set the [`default_vector_transport_method`](@extref `ManifoldsBase.default_vector_transport_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `vector_transport_method=` does not have to be specified.
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 * A [`get_basis`](@extref `ManifoldsBase.get_basis-Tuple{AbstractManifold, Any, ManifoldsBase.AbstractBasis}`) for the [`DefaultOrthonormalBasis`](@extref `ManifoldsBase.DefaultOrthonormalBasis`) on ``\mathcal M``
 * [`exp`](@extref `Base.exp-Tuple{AbstractManifold, Any, Any}`) and [`log`](@extref `Base.log-Tuple{AbstractManifold, Any, Any}`) (on ``\mathcal M``)
 * A [`DiagonalizingOrthonormalBasis`](@extref `ManifoldsBase.DiagonalizingOrthonormalBasis`) to compute the differentials of the exponential and logarithmic map

docs/src/solvers/truncated_conjugate_gradient_descent.md

@@ -51,7 +51,7 @@ The [`trust_regions`](@ref) solver requires the following functions of a manifol
 * if you do not provide a `trust_region_radius=`, then [`injectivity_radius`](@extref `ManifoldsBase.injectivity_radius-Tuple{AbstractManifold}`) on the manifold `M` is required.
 * the [`norm`](@extref `LinearAlgebra.norm-Tuple{AbstractManifold, Any, Any}`) as well, to stop when the norm of the gradient is small, but if you implemented `inner`, the norm is provided already.
 * A [`zero_vector!`](@extref `ManifoldsBase.zero_vector-Tuple{AbstractManifold, Any}`)`(M,X,p)`.
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 
 ## Literature
 

docs/src/solvers/trust_regions.md

@@ -66,7 +66,7 @@ The [`trust_regions`](@ref) solver requires the following functions of a manifol
 * A [`retract!`](@extref ManifoldsBase :doc:`retractions`)`(M, q, p, X)`; it is recommended to set the [`default_retraction_method`](@extref `ManifoldsBase.default_retraction_method-Tuple{AbstractManifold}`) to a favourite retraction. If this default is set, a `retraction_method=` does not have to be specified.
 * By default the stopping criterion uses the [`norm`](@extref `LinearAlgebra.norm-Tuple{AbstractManifold, Any, Any}`) as well, to stop when the norm of the gradient is small, but if you implemented `inner`, the norm is provided already.
 * if you do not provide an initial `max_trust_region_radius`, a [`manifold_dimension`](@extref `ManifoldsBase.manifold_dimension-Tuple{AbstractManifold}`) is required.
-* A [`copyto!](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
+* A [`copyto!`](@extref `Base.copyto!-Tuple{AbstractManifold, Any, Any}`)`(M, q, p)` and [`copy`](@extref `Base.copy-Tuple{AbstractManifold, Any}`)`(M,p)` for points.
 * By default the tangent vectors are initialized calling [`zero_vector`](@extref `ManifoldsBase.zero_vector-Tuple{AbstractManifold, Any}`)`(M,p)`.
 
 

src/Manopt.jl

@@ -11,7 +11,7 @@ module Manopt
 import Base: &, copy, getindex, identity, length, setindex!, show, |
 import LinearAlgebra: reflect!
 import ManifoldsBase: embed!, plot_slope, prepare_check_result, find_best_slope_window
-import ManifoldsBase: base_manifold, base_point
+import ManifoldsBase: base_manifold, base_point, get_basis
 using ColorSchemes
 using ColorTypes
 using Colors
@@ -364,6 +364,8 @@ export get_proximal_map, get_proximal_map!
 export get_state,
     get_initial_stepsize,
     get_iterate,
+    get_jacobian,
+    get_jacobian!,
     get_gradients,
     get_gradients!,
     get_manifold,
@@ -379,6 +381,8 @@ export get_state,
     get_differential_dual_prox!,
     set_gradient!,
     set_iterate!,
+    get_residuals,
+    get_residuals!,
     linearized_forward_operator,
     linearized_forward_operator!,
     adjoint_linearized_operator,