Some notes on continuity in convex optimization

Torbjørn Cunis

DOI: https://doi.org/10.48550/arXiv.2104.15045

2021-04-30

Optimization and Control

Abstract:We provide proof that the optimal value function of a convex parametrized optimization problem in Euclidean spaces is itself a convex function onto the extended real line.

What problem does this paper attempt to address?

Continuity of Optimal Solution Functions and their Conditions on Objective Functions

Yasushi Terazono,Ayumu Matani

DOI: https://doi.org/10.1137/110850189

IF: 2.763

2015-01-01

SIAM Journal on Optimization

Abstract:When an optimization problem has a set of parameters besides decision variables, the mapping from the parameters to the corresponding optimal solutions is determined accordingly. This mapping, either single-valued or set-valued, is called an optimal solution function in the literature of parametric optimization and a generalized inverse mapping in the literature of inverse problems. We present a pair of relationships between classes of objective functions and optimal solution functions in the form of two theorems. The first theorem states that any continuous quasi-convex objective function results in an upper semicontinuous optimal solution function. This is a variant of the Berge maximum theorem, wherein we replace the condition of compact-valuedness of the feasible solution function with other assumptions. The second theorem states that any optimal solution function with a certain continuity can be realized by a continuous convex objective function if the feasible solutions form an affine subspace. This is a variant of Komiya's inverse of Berge's theorem; we strengthened the assertion of convexity by assuming the costs to be nonparameterized, though we only consider the case of linear equation constraints.

mathematics, applied
Stable and Total Fenchel Duality for Convex Optimization Problems in Locally Convex Spaces

Chong Li,Donghui Fang,Genaro Lopez,Marco A. Lopez

DOI: https://doi.org/10.1137/080734352

IF: 2.763

2009-01-01

SIAM Journal on Optimization

Abstract:We consider the optimization problem $(P_{A})$ $\inf_{x\in X}\{f(x)+g(Ax)\}$ where $f$ and $g$ are proper convex functions defined on locally convex Hausdorff topological vector spaces $X$ and $Y$, respectively, and $A$ is a linear operator from $X$ to $Y$. By using the properties of the epigraph of the conjugated functions, some sufficient and necessary conditions for the strong Fenchel duality and the strong converse Fenchel duality of $(P_{A})$ are provided. Sufficient and necessary conditions for the stable Fenchel duality and for the total Fenchel duality are also derived.
Convergence analysis of the Gauss–Newton method for convex inclusion and convex-composite optimization problems

C. Li,K.F. Ng

DOI: https://doi.org/10.1016/j.jmaa.2011.11.062

IF: 1.417

2012-01-01

Journal of Mathematical Analysis and Applications

Abstract:Using the convex process theory we study the convergence issues of the iterative sequences generated by the Gauss–Newton method for the convex inclusion problem defined by a cone C and a smooth function F (the derivative is denoted by F′). The restriction in our consideration is minimal and, even in the classical case (the initial point x0 is assumed to satisfy the following two conditions: F′ is Lipschitz around x0 and the convex process Tx0, defined by Tx0⋅=F′(x0)⋅−C, is surjective), our results are new in giving sufficient conditions (which are weaker than the known ones and have a remarkable property being affine-invariant) ensuring the convergence of the iterative sequence with initial point x0. The same study is also made for the so-called convex-composite optimization problem (with objective function given as the composite of a convex function with a smooth map).
Continuity of Parametric Optima for Possibly Discontinuous Functions and Noncompact Decision Sets

Eugene A. Feinberg,Pavlo O. Kasyanov,David N. Kraemer

DOI: https://doi.org/10.48550/arXiv.2109.06299

2021-09-14

Abstract:This paper investigates continuity properties of value functions and solutions for parametric optimization problems. These problems are important in operations research, control, and economics because optimality equations are their particular cases. The classic fact, Berge's maximum theorem, gives sufficient conditions for continuity of value functions and upper semicontinuity of solution multifunctions. Berge's maximum theorem assumes that the objective function is continuous and the multifunction of feasible sets is compact-valued. These assumptions are not satisfied in many applied problems, which historically has limited the relevance of the theorem. This paper generalizes Berge's maximum theorem in three directions: (i) the objective function may not be continuous, (ii) the multifunction of feasible sets may not be compact-valued, and (iii) necessary and sufficient conditions are provided. To illustrate the main theorem, this paper provides applications to inventory control and to the analysis of robust optimization over possibly noncompact action sets and discontinuous objective functions.

Optimization and Control
Remarks on the continuity of convex functions in geodesic spaces

Genaro López-Acedo,Adriana Nicolae

DOI: https://doi.org/10.12775/tmna.2023.060

2024-04-04

Topological Methods in Nonlinear Analysis

Abstract:We explore the behavior of convex functions defined on locally polyhedral type spaces. As a relevant example of such spaces we show that locally compact -trees with the geodesic extension property around any point are locally polyhedral spaces.

mathematics
The Hadwiger Theorem on Convex Functions, I

Colesanti, Andrea

DOI: https://doi.org/10.1007/s00039-024-00693-8

2024-10-17

Geometric and Functional Analysis

Abstract:A complete classification of all continuous, epi-translation and rotation invariant valuations on the space of super-coercive convex functions on is established. The valuations obtained are functional versions of the classical intrinsic volumes. For their definition, singular Hessian valuations are introduced.

mathematics
A note on continuity and consistency of measures of risk and variability

Niushan Gao,Foivos Xanthos

2024-05-16

Abstract:In this short note, we show that every convex, order bounded above functional on a Banach lattice is automatically norm continuous. This improves a result in \cite{RS06} and applies to many deviation and variability measures. We also show that an order-continuous, law-invariant functional on an Orlicz space is strongly consistent everywhere, extending a result in \cite{KSZ14}.

Risk Management
Equivariant Valuations on Convex Functions

Georg C. Hofstätter,Jonas Knoerr

2024-07-11

Abstract:We classify all continuous valuations on the space of finite convex functions with values in the same space which are dually epi-translation-invariant and equi- resp. contravariant with respect to volume-preserving linear maps. We thereby identify the valuation-theoretic functional analogues of the difference body map and show that there does not exist a generalization of the projection body map in this setting. This non-existence result is shown to also hold true for valuations with values in the space of convex functions that are finite in a neighborhood of the origin.

Metric Geometry,Functional Analysis
On existence of solutions to non-convex minimization problems

Rohan Rele,Angelia Nedich

2024-05-08

Abstract:We provide new sufficient conditions for the finiteness of the optimal value and existence of solutions to a general problem of minimizing a proper closed function over a nonempty closed set. The conditions require an asymptotically bounded decay of a function, a relaxation of p-supercoercivity, and a certain relation for the asymptotic cone of the constraint set and the asymptotic function of the objective function. Our analysis combines these conditions with a regularization technique. We refine the notion of retractive directions of a set, extend its definition to functions, and establish some basic relations for such directions for both sets and functions. Using these tools, we provide existence of solutions results that generalize many of the results in the literature for both non-convex and convex problems.

Optimization and Control
Continuity of the value function in sparse optimal control

Takuya Ikeda,Masaaki Nagahara

DOI: https://doi.org/10.1109/ascc.2015.7244652

2015-05-01

Abstract:The purpose of this article is to show the continuity of the value function of the sparse optimal (or $L^{\mathrm{0}}$-optimal) control problem. The sparse optimal control is a control whose support is minimum among all admissible controls. Under the normality assumption, it is known that a sparse optimal control is given by $L^{1}$ optimal control. Furthermore, the value function of the sparse optimal control problem is identical with that of the $L^{1}$-optimal control problem. From these properties, we prove the continuity of the value function of the sparse optimal control problem by verifying that of the $L^{1}$-optimal control problem.
Convexity and Osculation in Normed Spaces

Mark Mandelkern

DOI: https://doi.org/10.1216/RMJ-2013-43-2-551

2024-04-04

Abstract:Constructive properties of uniform convexity, strict convexity, near convexity, and metric convexity in real normed linear spaces are considered. Examples show that certain classical theorems, such as the existence of points of osculation, are constructively invalid. The methods used are in accord with principles introduced by Errett Bishop

Functional Analysis
Continuous Convexity Measures

Abel Douzal,Ferdinand Jacobé de Naurois

2023-06-03

Abstract:Methods for measuring convexity defects of compacts in R^n abound. However, none of the those measures seems to take into account continuity. Continuity in convexity measure is essential for optimization, stability analysis, global optimality, convergence analysis, and accurate modelling as it ensures robustness and facilitates the development of efficient algorithms for solving convex optimization problems. This paper revisits the axioms underlying convexity measures by enriching them with a continuity hypothesis in Hausdorff's sense. Having provided the concept's theoretical grounds we state a theorem underlining the necessity of restricting ourselves to non-point compacts. We then construct a continuous convexity measure and compare it to existing measures.

Geometric Topology,Optimization and Control
Continuity of extensions of Lipschitz maps and of monotone maps

Krzysztof J. Ciosmak

DOI: https://doi.org/10.1112/jlms.70014

2024-11-04

Journal of the London Mathematical Society

Abstract:Abstract Let be a subset of a Hilbert space. We prove that if is such that for all and all non‐negative that add up to one, then for any 1‐Lipschitz , with , there exists a 1‐Lipschitz extension of such that the uniform distance on between and is the same as the uniform distance on between and . Moreover, if either or is convex, we prove the converse: We show that a map that allows for a 1‐Lipschitz, uniform distance preserving extension of any 1‐Lipschitz map on a subset of also satisfies the above set of inequalities. We also prove a similar continuity result concerning extensions of monotone maps. Our results hold true also for maps taking values in infinite‐dimensional spaces.

mathematics
On semi-continuity and continuity of the smallest and largest minimizing point of real convex functions with applications in probability and statistics

Dietmar Ferger

2023-11-05

Abstract:We prove that the smallest minimizer s(f) of a real convex function f is less than or equal to a real point x if and only if the right derivative of f at x is non-negative. Similarly, the largest minimizer t(f) is greater or equal to x if and only if the left derivative of f at x is non-positive. From this simple result we deduce measurability and semi-continuity of the functionals s and t. Furthermore, if f has a unique minimizing point, so that s(f) = t(f), then the functional is continuous at f. With these analytical preparations we can apply Continuous Mapping Theorems to obtain several Argmin theorems for convex stochastic processes. The novelty here are statements about classical distributional convergence and almost sure convergence, if the limit process does not have a unique minimum point. This is possible by replacing the natural topology on R with the order topologies. Another new feature is that not only sequences but more generally nets of convex stochastic processes are allowed.

Probability
Marginal Analysis of Convex Optimization Problems with Set-Valued Inclusion Constraints

Amos Uderzo

2024-12-05

Abstract:In this paper, stability and sensitivity properties of a class of parametric constrained optimization problem, whose feasible region is defined by a set-valued inclusion, are investigated through the associated optimal value function. Set-valued inclusions are a kind of constraint system, which naturally emerges in contexts requiring the robust fulfilment of traditional cone constraints, where data are affected by uncertain elements having a non stochastic nature, or in (MPEC) as a vector equilibrium constraint, where feasible solutions are intended as equilibrium point in a strong sense. Under proper convexity assumptions on the objective function and the constraining set-valued term, combined with a global qualification condition, a class of parametric optimization problems is singled out, which displays a global Lipschitz behaviour. By employing recent results of variational analysis, elements for a sensitivity analysis of this class of problems are provided via exact subgradient formulae for the optimal value function. Further consequences of the stability behaviour are explored in terms of problem calmness and viability of penalization techniques.

Optimization and Control
On continuation and convex Lyapunov functions

Wouter Jongeneel,Roland Schwan

2024-03-21

Abstract:Suppose that the origin is globally asymptotically stable under a set of continuous vector fields on Euclidean space and suppose that all those vector fields come equipped with -- possibly different -- convex Lyapunov functions. We show that this implies there is a homotopy between any two of those vector fields such that the origin remains globally asymptotically stable along the homotopy. Relaxing the assumption on the origin to any compact convex set or relaxing convexity to geodesic convexity does not alter the conclusion. Imposing the same convexity assumptions on control Lyapunov functions leads to a Hautus-like stabilizability test. These results ought to be of interest in the context of learning stability certificates, policy gradient methods and switched systems.

Optimization and Control,Dynamical Systems
On the Relationship Between the Integer and Continuous Solutions of Convex Programs.

XL Sun,D Li

DOI: https://doi.org/10.1016/s0167-6377(01)00082-7

IF: 1.151

2001-01-01

Operations Research Letters

Abstract:A bound is obtained in this note for the distance between the integer and real solutions to convex quadratic programs. This bound is a function of the condition number of the Hessian matrix. We further extend this proximity result to convex programs and mixed-integer convex programs. We also show that this bound is achievable in certain situations and the distance between the integer and continuous minimizers may tend to infinity.
About the continuity of one operation with convex compacts in finite-dimensional normed spaces

A. Kh. Galstyan

DOI: https://doi.org/10.48550/arXiv.2211.03847

2022-11-08

Abstract:In this paper, we study the deformation of the intersection of one compact set with a closed neighborhood of another compact set by changing the radius of this neighborhood. It is shown that in finite-dimensional normed spaces, in the case when both compact sets are non-empty convex subsets, such an operation is continuous in the topology generated by the Hausdorff metric. The question of the continuous dependence of the described intersection on the radius of the neighborhood arose as a by-product of the development of the theory of extremal networks. However, it turned out to be interesting in itself, suggesting various generalizations. Therefore, it was decided to publish it separately.

Metric Geometry
On the Lebesgue Property of Monotone Convex Functions

Keita Owari

DOI: https://doi.org/10.48550/arXiv.1305.2271

2013-05-10

Functional Analysis

Abstract:The Lebesgue property (order-continuity) of a monotone convex function on a solid vector space of measurable functions is characterized in terms of (1) the weak inf-compactness of the conjugate function on the order-continuous dual space, (2) the attainment of the supremum in the dual representation by order-continuous linear functionals. This generalizes and unifies several recent results obtained in the context of convex risk measures.

Some notes on continuity in convex optimization

Continuity of Optimal Solution Functions and their Conditions on Objective Functions

Stable and Total Fenchel Duality for Convex Optimization Problems in Locally Convex Spaces

Convergence analysis of the Gauss–Newton method for convex inclusion and convex-composite optimization problems

Continuity of Parametric Optima for Possibly Discontinuous Functions and Noncompact Decision Sets

Remarks on the continuity of convex functions in geodesic spaces

The Hadwiger Theorem on Convex Functions, I

A note on continuity and consistency of measures of risk and variability

Equivariant Valuations on Convex Functions

On existence of solutions to non-convex minimization problems

Continuity of the value function in sparse optimal control

Convexity and Osculation in Normed Spaces

Continuous Convexity Measures

Continuity of extensions of Lipschitz maps and of monotone maps

On semi-continuity and continuity of the smallest and largest minimizing point of real convex functions with applications in probability and statistics

Marginal Analysis of Convex Optimization Problems with Set-Valued Inclusion Constraints

On continuation and convex Lyapunov functions

On the Relationship Between the Integer and Continuous Solutions of Convex Programs.

About the continuity of one operation with convex compacts in finite-dimensional normed spaces

On the Lebesgue Property of Monotone Convex Functions