AI RESEARCH
On the Communication Complexity of Decentralized Stochastic Bilevel Optimization
arXiv CS.LG
•
ArXi:2311.11342v5 Announce Type: replace Stochastic bilevel optimization finds widespread applications in machine learning, including meta-learning, hyperparameter optimization, and neural architecture search. To extend stochastic bilevel optimization to distributed data, several decentralized stochastic bilevel optimization algorithms have been developed. However, existing methods often suffer from slow convergence rates and high communication costs in heterogeneous settings, limiting their applicability to real-world tasks.