9 problems

For every $t\in\mathbb{N}$, define the reactivation-gap contribution … These gaps arise from components updated more than once between iterations $t$ and $t+K-1$. Improved-converge…

Let $d_{\max}$ denote the maximum depth of recursion in the recursive Frank–Wolfe algorithms, and let the runtime of their optimization oracles be measured as a function of the ins…

Let a smooth, strongly convex objective be minimized over a convex and compact polytope. The Frank–Wolfe method is augmented with away-steps, which move away from previously found…

Abernethy et al.'s conjecture. $K$ is a gauge set with respect to its gauge $\|\cdot\|_K$ if and only if it is strongly convex with respect to $K$ itself.

Away-mechanism trade-off conjecture. The away mechanism improves the trade-off between the cost of the LMO and the alignment of the descent direction with the opposite of the gradi…

Consider general uniformly strongly convex-concave functions in both cases described by the source as situations … . Linear-rate conjecture. A refined analysis could yield a linear…

Let the SP-FW algorithm use step sizes $\gamma_t=\nicefrac{1}{(t+1)}$, and let it be applied to a uniformly strongly convex-concave objective function over a polytope. Hammond's co…

Let $\mathcal{A}$ be a set of vertices, meaning points that are extrema of their convex hull, and suppose that a new vertex is added while all previous points remain vertices. Pyra…

Norm-selection conjecture. In general, the flexibility to choose the norm should make the ratios of these boundedness constants to $C_f^\tau$ well controlled, while the typical dep…