arXiv cs.AI / cs.LG / cs.CL·15d agoQuantile-based Loss Filtering for Outlier-Robust Stochastic Gradient Descent#convergence#optimization#outlier-robustnessAI research