Mathematics > Optimization and Control
[Submitted on 27 Sep 2024]
Title:Concave tents: a new tool for constructing concave reformulations of a large class of nonconvex optimization problems
View PDFAbstract:Optimizing a nonlinear function over nonconvex sets is challenging since solving convex relaxations may lead to substantial relaxation gaps and infeasible solutions, that must be "rounded" to feasible ones, often with uncontrollable losses in objective function performance. For this reason, these convex hulls are especially useful if the objective function is linear or even concave since concave optimization is invariant to taking the convex hull of the feasible set. Motivated by this observation, we propose the notion of concave tents, which are concave approximations of the original objective function that agree with this objective function on the feasible set, and allow for concave reformulations of the problem. We derive these concave tents for a large class of objective functions as the optimal value functions of conic optimization problems. Hence, evaluating our concave tents requires solving a conic problem. Interestingly, we can find supergradients by solving the conic dual problem, so that differentiation is of the same complexity as evaluation. For feasible sets that are contained in the extreme points of their convex hull, we construct these concave tents in the original space of variables. For general feasible sets, we propose a double lifting strategy, where the original optimization problem is lifted into a higher dimensional space in which the concave tent can be constructed with a similar effort. We investigate the relation of the so-constructed concave tents to concave envelopes and a naive concave tent based on concave quadratic updates. Based on these ideas we propose a primal heuristic for a class of robust discrete quadratic optimization problems, that can be used instead of classical rounding techniques. Numerical experiments suggest that our techniques can be beneficial as an upper bounding procedure in a branch and bound solution scheme.
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.