Revisiting the Approximate Carathéodory Problem via the Frank-Wolfe Algorithm [article]

Cyrille W. Combettes, Sebastian Pokutta
2021 arXiv   pre-print
The approximate Carathéodory theorem states that given a compact convex set 𝒞⊂ℝ^n and p∈[2,+∞[, each point x^*∈𝒞 can be approximated to ϵ-accuracy in the ℓ_p-norm as the convex combination of 𝒪(pD_p^2/ϵ^2) vertices of 𝒞, where D_p is the diameter of 𝒞 in the ℓ_p-norm. A solution satisfying these properties can be built using probabilistic arguments or by applying mirror descent to the dual problem. We revisit the approximate Carathéodory problem by solving the primal problem via the Frank-Wolfe
more » ... algorithm, providing a simplified analysis and leading to an efficient practical method. Furthermore, improved cardinality bounds are derived naturally using existing convergence rates of the Frank-Wolfe algorithm in different scenarios, when x^* is in the interior of 𝒞, when x^* is the convex combination of a subset of vertices with small diameter, or when 𝒞 is uniformly convex. We also propose cardinality bounds when p∈[1,2[∪{+∞} via a nonsmooth variant of the algorithm. Lastly, we address the problem of finding sparse approximate projections onto 𝒞 in the ℓ_p-norm, p∈[1,+∞].
arXiv:1911.04415v5 fatcat:ymeeb5d4bjezre7cw43rn3vyuy