A proximal subgradient algorithm with extrapolation for structured nonconvex nonsmooth problems (Q6141533)

In the paper under review, the authors study a following broad optimization problem which has many important applications in diverse areas, including power control problems, compressed sensing, portfolio optimization, supply chain problems, image segmentation and many others. The objective function is formed by the sum of a possibly nonsmooth nonconvex function and a differentiable function with Lipschitz continuous gradient, subtracted by a weakly convex function. This general framework allows for problems involving nonconvex loss functions and problems with specific nonconvex constraints. The problem is: \(\min_{x\in C} F(x)\) where \(F(x) := f (x) + h(Ax)-g(x)\) and where \(C\) is a nonempty closed subset of a finite-dimensional real Hilbert space \(H\), \(A\) is a linear mapping from \(H\) to another finite-dimensional real Hilbert space, \(f : H\to (-\infty,\infty]\) is a proper lower semicontinuous (possibly nonsmooth and nonconvex) function, \(h\) is a real valued differentiable (possibly nonconvex) function whose gradient is Lipschitz continuous and \(g:H\to (-\infty,\infty]\) is a continuous weakly convex function with modulus on an open convex set containing \(C\). Two examples are: (1) From statistical learning, \(\min_{x\in \mathbb R^d}(\phi(x) + \gamma r(x))\) where \(\phi\) is called a loss function which measures the data fitting, \(r\) is a regularization which promotes specific structure in the solution such as sparsity, and \(\gamma>0\) is a weighting parameter. (2) \(\min_{x\in \mathbb R^d} (f (x) + h(x)- g(x))\). The paper is well written with a good set of references.