Researchers have developed novel variance-reduction techniques for stochastic composite inclusions, introducing both unbiased and biased estimators. The unbiased methods, including mini-batch SGD and loopless-SVRG, achieve an $\mathcal{O}(1/k)$ convergence rate. The newly introduced biased estimators, such as SARAH and Hybrid SGD, also maintain convergence while offering different oracle complexities. These methods were demonstrated through numerical experiments in AUC optimization for imbalanced classification and policy evaluation in reinforcement learning. AI
IMPACT Introduces new optimization techniques that could improve the efficiency of training machine learning models, particularly in areas like imbalanced classification and reinforcement learning.
RANK_REASON The cluster contains an academic paper detailing new methods and theoretical results in machine learning. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
