On strongly quasiconvex functions: existence results and proximal point algorithms (Q2116608)

A function \(f:X\to \mathbb{R}\) is \(\gamma\)-strongly quasiconvex if for any \(x,y\in X\) and \(\lambda\in [0,1]\), \[f(\lambda x + (1-\lambda)y)\leq \max(f(x),f(y)) - \gamma \|x-y\|^2.\] The purpose of this paper is to study these functions and their proximal operators. Namely, it is shown that strongly quasiconvex functions are 2-supercoercive (that is, \(\liminf_{\|x\|\to +\infty} h(x)/\|x\|^2 > 0\)), and therefore coercive, and that when they are lsc they have a unique minimizer. The proximal operator of strongly quasiconvex functions also enjoys important properties, in particular the fixed points of the proximal operators of a proper strongly quasiconvex function \(h\) are the minimizers of that function. As a consequence of this, it is shown that the classical proximal point algorithm (where \(x^{k+1}\in\mathrm{Prox}_{c_kh}(K,x^k)\)) converges whenever the sequence \(c_k\) is positive and bounded away from 0. A proximal point algorithm is also given for quasiconvex functions based on these results.