MARTHE: Scheduling the Learning Rate Via Online Hypergradients

October 18, 2019 · Entered Twilight · 🏛 arXiv.org

"Last commit was 5.0 years ago (≥5 year threshold)"

Evidence collected by the PWNC Scanner

Repo contents: CODE_OF_CONDUCT.md, CONTRIBUTING.md, LICENSE.txt, NOTICE, README.md, VERSION, adatune, bin, figures, setup.cfg, setup.py

Authors Michele Donini, Luca Franceschi, Massimiliano Pontil, Orchid Majumder, Paolo Frasconi arXiv ID 1910.08525 Category cs.LG: Machine Learning Cross-listed stat.ML Citations 7 Venue arXiv.org Repository https://github.com/awslabs/adatune ⭐ 291 Last Checked 4 months ago

Abstract

We study the problem of fitting task-specific learning rate schedules from the perspective of hyperparameter optimization, aiming at good generalization. We describe the structure of the gradient of a validation error w.r.t. the learning rate schedule -- the hypergradient. Based on this, we introduce MARTHE, a novel online algorithm guided by cheap approximations of the hypergradient that uses past information from the optimization trajectory to simulate future behaviour. It interpolates between two recent techniques, RTHO (Franceschi et al., 2017) and HD (Baydin et al. 2018), and is able to produce learning rate schedules that are more stable leading to models that generalize better.