SLIP is the first single-loop stochastic bilevel optimizer with eO(1/epsilon^4) oracle complexity under unbounded upper-level smoothness, both in expectation and with high probability.
Coresets via bilevel optimization for continual learning and streaming
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Nearly Optimal Single Loop Algorithm for Stochastic Bilevel Optimization under Unbounded Smoothness
SLIP is the first single-loop stochastic bilevel optimizer with eO(1/epsilon^4) oracle complexity under unbounded upper-level smoothness, both in expectation and with high probability.