STORM+: Fully Adaptive SGD with Momentum for Nonconvex Optimization
6. prosince 2021
Řečníci
Kfir Y. Levy
Sprecher:in · 0 Follower:innen
Volkan Cevher
Sprecher:in · 0 Follower:innen
O prezentaci
In this work we investigate stochastic non-convex optimization problems where the objective is an expectation over smooth loss functions, and the goal is to find an approximate stationary point. The most popular approach to handling such problems is variance reduction techniques, which are also known to obtain tight convergence rates, matching the lower bounds in this case. Nevertheless, these techniques require a careful maintenance of anchor points in conjunction with appropriately selected “`"mega-batchsizes". This leads to a challenging hyperparameter tuning problem, that weakens their practicality. Recently, [Cutkosky & Orabona, 2019] have shown that one can employ recursive momentum in order to avoid the use of anchor points and large batchsizes, and still obtain the optimal rate for this setting. Yet, their method called STORM crucially relies on the knowledge of the smoothness, as well a bound on the gradient norms. We propose STORM^+, a new method that is completely parameter-free, does not require large batch-sizes, and obtains the optimal O(1/T^1/3) rate for finding an approximate stationary point. Our work builds on the STORM algorithm, in conjunction with a novel approach to adaptively set the learning rate and momentum parameters.In this work we investigate stochastic non-convex optimization problems where the objective is an expectation over smooth loss functions, and the goal is to find an approximate stationary point. The most popular approach to handling such problems is variance reduction techniques, which are also known to obtain tight convergence rates, matching the lower bounds in this case. Nevertheless, these techniques require a careful maintenance of anchor points in conjunction with appropriately selected “…
Organizátor
NeurIPS 2021
Konto · 1,9k Follower:innen
O organizátorovi (NeurIPS 2021)
Neural Information Processing Systems (NeurIPS) is a multi-track machine learning and computational neuroscience conference that includes invited talks, demonstrations, symposia and oral and poster presentations of refereed papers. Following the conference, there are workshops which provide a less formal setting.
Baví vás formát? Nechte SlidesLive zachytit svou akci!
Profesionální natáčení a streamování po celém světě.
Sdílení
Doporučená videa
Prezentace na podobné téma, kategorii nebo přednášejícího
ToAlign: Task-Oriented Alignment for Unsupervised Domain Adaptation
Später ansehen
Guoqiang Wei, …
Ewigspeicher-Fortschrittswert: 0 = 0.0%
AI workloads inside databases
Später ansehen
Ewigspeicher-Fortschrittswert: 0 = 0.0%
2nd Workshop on Self-Supervised Learning: Theory and Practice
Später ansehen
Ewigspeicher-Fortschrittswert: 0 = 0.0%
Fast Axiomatic Attribution for Neural Networks
Später ansehen
Robin Hesse, …
Ewigspeicher-Fortschrittswert: 0 = 0.0%
Confident Anchor-Induced Multi-Source Free Domain Adaptation
Später ansehen
Ewigspeicher-Fortschrittswert: 0 = 0.0%
Machine Learning for Healthcare
Später ansehen
Ewigspeicher-Fortschrittswert: 0 = 0.0%