-
arXiv:2506.23303 [pdf, ps, other]
On the boundedness of the sequence generated by minibatch stochastic gradient descent
Abstract: Stochastic Gradient Descent (SGD) with Polyak's stepsize has recently gained renewed attention in stochastic optimization. Recently, Orvieto, Lacoste-Julien, and Loizou introduced a decreasing variant of Polyak's stepsize, where convergence relies on a boundedness assumption of the iterates. They established that this assumption holds under strong convexity. In this paper, we extend their result b… ▽ More
Submitted 29 June, 2025; originally announced June 2025.
MSC Class: Primary 90C15; 90C25; 65K05; Secondary 68T07; 68W20; 68W40
-
arXiv:2506.22553 [pdf, ps, other]
On a result by Meshulam
Abstract: In 1996, Meshulam proved that every sequence generated by applying projections onto affine subspaces, drawn from a finite collection in Euclidean space, must be bounded. In this paper, we extend his result not only from affine subspaces to convex polyhedral subsets, but also from Euclidean to general Hilbert space. Various examples are provided to illustrate the sharpness of the results.
Submitted 27 June, 2025; originally announced June 2025.
MSC Class: 47H09; 65K05; 90C25 (Primary) 52A37; 52B55 (Secondary)