Amélie Lambert
Research Subjects
Amélie Lambert has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
Sourour Elloumi
6 shared publications- Alain Billionnet
Laboratoire CEDRIC, École Nationale Supérieure d’Informatique pour l’Industrie et l’Entreprise
3 shared publications - Daniel Porumbel
CEDRIC-CNAM, Paris Cedex 03, France
3 shared publications - Sourour Elloumi
ENSIIE-CEDRIC 1 square de la résistance FR-91025 Evry
3 shared publications - Sourour Elloumi
UMA, ENSTA, Institut Polytechnique de Paris, 91120 Palaiseau, France
3 shared publications Alain Billionnet
CEDRIC-ENSIIE
2 shared publications
Latest Publications
- 5 September 2026
- Published by Springer Nature
- p. 132-149
Fast SDP Certification of Neural Networks : Towards Large Multi-class Datasets
- 5 September 2026
- Published by Springer Nature
- p. 132-149
Fast SDP Certification of Neural Networks : Towards Large Multi-class Datasets
- Open Access
Polynomial Optimization by Quadratization and Convexification
The question of minimizing a box-constrained polynomial (in mixed-integer variables) can easily and frequently emerge across many fields of science and engineering. This work explores the idea of solving this problem by reformulating the given polynomial as a quadratic function. Such reformulation induces a number of auxiliary variables that are intended to represent various (high-degree) monomials in the original variables. The interdependence of the artificial and the original variables is enforced through a hierarchical collection of non-convex quadratic constraints, referred to as a quadratization scheme. We thus obtain a quadratically-constrained quadratic program (QCQP) that is completely equivalent to the given polynomial program. This QCQP is still non-convex and N P-hard, but it can be relaxed into a convex quadratic program more easily than an arbitrary-degree polynomial. It is easier to convexify a quadratic function since its Hessian matrix does not vary with the point of evaluation as in the case of a higher-degree polynomial. The coefficients of the resulting convex quadratic program are obtained from the optimal dual values of the SDP (positive semi-definite) relaxation of the same QCQP program. We further integrate it into a spatial Branch-and-bound framework, in which each branching node reduces to a box that simply limits the domain of each variable. The Branch-and-bound is powered by the above convex quadratic program: the lower bound of each node is determined by solving this convex relaxation over the corresponding box, which is a very fast operation using modern Interior Point Methods solvers. Numerical results confirm the potential of the proposed ideas.
Open AccessPolynomial Optimization by Quadratization and Convexification
The question of minimizing a box-constrained polynomial (in mixed-integer variables) can easily and frequently emerge across many fields of science and engineering. This work explores the idea of solving this problem by reformulating the given polynomial as a quadratic function. Such reformulation induces a number of auxiliary variables that are intended to represent various (high-degree) monomials in the original variables. The interdependence of the artificial and the original variables is enforced through a hierarchical collection of non-convex quadratic constraints, referred to as a quadratization scheme. We thus obtain a quadratically-constrained quadratic program (QCQP) that is completely equivalent to the given polynomial program. This QCQP is still non-convex and N P-hard, but it can be relaxed into a convex quadratic program more easily than an arbitrary-degree polynomial. It is easier to convexify a quadratic function since its Hessian matrix does not vary with the point of evaluation as in the case of a higher-degree polynomial. The coefficients of the resulting convex quadratic program are obtained from the optimal dual values of the SDP (positive semi-definite) relaxation of the same QCQP program. We further integrate it into a spatial Branch-and-bound framework, in which each branching node reduces to a box that simply limits the domain of each variable. The Branch-and-bound is powered by the above convex quadratic program: the lower bound of each node is determined by solving this convex relaxation over the corresponding box, which is a very fast operation using modern Interior Point Methods solvers. Numerical results confirm the potential of the proposed ideas.
- 5 October 2025
- Published by Springer Nature in Journal of Combinatorial Optimization
- Vol. 50 (3) , 1-47
Quadratization and convexification in polynomial binary optimization
- 5 October 2025
- Published by Springer Nature in Journal of Combinatorial Optimization
- Vol. 50 (3) , 1-47
Quadratization and convexification in polynomial binary optimization
- 30 June 2025
- Published by Springer Nature in Journal of Global Optimization
- Vol. 94 (2) , 399-425
Using quadratic cuts to iteratively strengthen convexifications of box quadratic programs
- 30 June 2025
- Published by Springer Nature in Journal of Global Optimization
- Vol. 94 (2) , 399-425
Using quadratic cuts to iteratively strengthen convexifications of box quadratic programs
- 22 May 2025
- Published by Springer Nature
- p. 451-463
Interval Observer for Battery and Motor Circuit Model of Electric Vehicles
- 22 May 2025
- Published by Springer Nature
- p. 451-463
Interval Observer for Battery and Motor Circuit Model of Electric Vehicles
- Open Access
Using quadratic cuts to iteratively strengthen convexifications of box quadratic programs
Published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Using quadratic cuts to iteratively strengthen convexifications of box quadratic programs
Open AccessUsing quadratic cuts to iteratively strengthen convexifications of box quadratic programs
Published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Using quadratic cuts to iteratively strengthen convexifications of box quadratic programs
- Open Access
- 26 March 2024
- Published by Elsevier in Computers & Operations Research
- Vol. 166, 106626
A tight compact quadratically constrained convex relaxation of the Optimal Power Flow problem
- 26 March 2024
- Published by Elsevier in Computers & Operations Research
- Vol. 166, 106626
Open AccessA tight compact quadratically constrained convex relaxation of the Optimal Power Flow problem
- 12 February 2024
- Published by Springer Nature in Journal of Global Optimization
- Vol. 91 (2) , 331-353
Global solution of quadratic problems using interval methods and convex relaxations
- 12 February 2024
- Published by Springer Nature in Journal of Global Optimization
- Vol. 91 (2) , 331-353
Global solution of quadratic problems using interval methods and convex relaxations
- 16 January 2023
- Published by Taylor & Francis in Optimization Methods and Software
- Vol. 38 (3) , 626-653
Using general triangle inequalities within quadratic convex reformulation method
We consider the exact solution of Problem (P) which consists in minimizing a quadratic function subject to quadratic constraints. We start with an explicit description of new general triangle inequalities that are derived from the ranges of the variables of (P). We show that they extend the triangle inequalities, introduced for the binary case, to variables that belong to a generic interval. We also prove that these inequalities cut feasible solutions of McCormick envelopes, and we relate them to the literature. We then introduce (SDP), a strong semidefinite relaxation of (P), that we call ‘Shor's plus RLT plus Triangle’, which includes both the McCormick envelopes and the general triangle inequalities. We further show how to compute a convex relaxation whose optimal value reaches the value of (SDP). In order to handle these inequalities in the solution of (SDP), we solve it by a heuristic that also serves as a separation algorithm. We then solve (P) to global optimality with a branch-and-bound based on . Finally, we show that our method outperforms the compared solvers.- 16 January 2023
- Published by Taylor & Francis in Optimization Methods and Software
- Vol. 38 (3) , 626-653
Using general triangle inequalities within quadratic convex reformulation method
We consider the exact solution of Problem (P) which consists in minimizing a quadratic function subject to quadratic constraints. We start with an explicit description of new general triangle inequalities that are derived from the ranges of the variables of (P). We show that they extend the triangle inequalities, introduced for the binary case, to variables that belong to a generic interval. We also prove that these inequalities cut feasible solutions of McCormick envelopes, and we relate them to the literature. We then introduce (SDP), a strong semidefinite relaxation of (P), that we call ‘Shor's plus RLT plus Triangle’, which includes both the McCormick envelopes and the general triangle inequalities. We further show how to compute a convex relaxation whose optimal value reaches the value of (SDP). In order to handle these inequalities in the solution of (SDP), we solve it by a heuristic that also serves as a separation algorithm. We then solve (P) to global optimality with a branch-and-bound based on . Finally, we show that our method outperforms the compared solvers.- Open Access
Quadratization and convexification in polynomial binary optimization
In this paper, we discuss several reformulations and solution approaches for the problem of minimizing a polynomial in binary variables (P). We review and integrate different literature streams to describe a methodology consisting of three distinct phases, together with several possible variants for each phase. The first phase determines a recursive decomposition of each monomial of interest into pairs of submonomials, down to the initial variables. The decomposition gives rise to a so-called quadratization scheme. The second phase builds a quadratic reformulation of (P) from a given quadratization scheme, by associating a new auxiliary variable with each submonomial that appears in the scheme. A quadratic reformulation of (P) is obtained by enforcing relations between the auxiliary variables and the monomials that they represent, either through linear constraints or through penalty terms in the objective function. The resulting quadratic problem (QP) is non-convex in general and is still difficult to solve. At this stage we introduce the third phase of the resolution process, which consists in convexifying (QP). We consider different types of convexification methods, including complete linearization or quadratic convex reformulations. Theoretical properties of the different phases are recalled from the literature or are further clarified. Finally, we present some experimental results to illustrate the discussion.
Open AccessQuadratization and convexification in polynomial binary optimization
In this paper, we discuss several reformulations and solution approaches for the problem of minimizing a polynomial in binary variables (P). We review and integrate different literature streams to describe a methodology consisting of three distinct phases, together with several possible variants for each phase. The first phase determines a recursive decomposition of each monomial of interest into pairs of submonomials, down to the initial variables. The decomposition gives rise to a so-called quadratization scheme. The second phase builds a quadratic reformulation of (P) from a given quadratization scheme, by associating a new auxiliary variable with each submonomial that appears in the scheme. A quadratic reformulation of (P) is obtained by enforcing relations between the auxiliary variables and the monomials that they represent, either through linear constraints or through penalty terms in the objective function. The resulting quadratic problem (QP) is non-convex in general and is still difficult to solve. At this stage we introduce the third phase of the resolution process, which consists in convexifying (QP). We consider different types of convexification methods, including complete linearization or quadratic convex reformulations. Theoretical properties of the different phases are recalled from the literature or are further clarified. Finally, we present some experimental results to illustrate the discussion.