S-Step Methods on a.devhttps://aditya08.github.io/tags/s-step-methods/Recent content in S-Step Methods on a.devHugo -- 0.148.0enThu, 23 Jan 2025 00:00:00 +0000Scalable Dual Coordinate Descent for Kernel Methodshttps://aditya08.github.io/publications/kerneldcd/Thu, 23 Jan 2025 00:00:00 +0000https://aditya08.github.io/publications/kerneldcd/This paper presents communication-avoiding, s-step dual coordinate descent methods for kernel SVM and kernel ridge regression. Winner of the HPC Asia 2025 Outstanding Paper Award.Communication-Efficient, 2D Parallel Stochastic Gradient Descent for Distributed-Memory Optimizationhttps://aditya08.github.io/publications/hybridsgd/Mon, 13 Jan 2025 00:00:00 +0000https://aditya08.github.io/publications/hybridsgd/This paper presents HybridSGD, a 2D parallel stochastic gradient descent method that trades off communication between s-step SGD and Federated Averaging