Sharp statistically optimal recovery guarantees for second-order critical points of nonconvex matrix LASSO under RIP, with counterexamples showing overparametrization does not always improve the landscape.
LoRA training provably converges to a low-rank global minimum or it fails loudly (but it probably won’t fail),
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Sharp recovery and landscape guarantees for the nonconvex matrix LASSO
Sharp statistically optimal recovery guarantees for second-order critical points of nonconvex matrix LASSO under RIP, with counterexamples showing overparametrization does not always improve the landscape.