1 problem

0 votes0 replies0 views
The general first-order stochastic optimization lower-bound conjecture under $\ell_1$ gradient measurement
Let a general first-order stochastic optimization algorithm optimize an objective under the assumptions associated with the stated SGD lower bound, and measure gradients by the…
Conjectureopengradient-norms optimization-lower-bounds stochastic-optimizationarXiv posted 2 days ago

The general first-order stochastic optimization lower-bound conjecture under ℓ1\ell_1ℓ1​ gradient measurement