We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Perturbated Gradients Updating within Unit Space for Deep Learning

Abstract: In deep learning, optimization plays a vital role. By focusing on image classification, this work investigates the pros and cons of the widely used optimizers, and proposes a new optimizer: Perturbated Unit Gradient Descent (PUGD) algorithm with extending normalized gradient operation in tensor within perturbation to update in unit space. Via a set of experiments and analyses, we show that PUGD is locally bounded updating, which means the updating from time to time is controlled. On the other hand, PUGD can push models to a flat minimum, where the error remains approximately constant, not only because of the nature of avoiding stationary points in gradient normalization but also by scanning sharpness in the unit ball. From a series of rigorous experiments, PUGD helps models to gain a state-of-the-art Top-1 accuracy in Tiny ImageNet and competitive performances in CIFAR- {10, 100}. We open-source our code at link: this https URL
Subjects: Machine Learning (cs.LG)
Cite as: arXiv:2110.00199 [cs.LG]
  (or arXiv:2110.00199v2 [cs.LG] for this version)

Submission history

From: Ching-Hsun Tseng [view email]
[v1] Fri, 1 Oct 2021 04:00:51 GMT (792kb)
[v2] Mon, 24 Jan 2022 18:25:25 GMT (9672kb)

Link back to: arXiv, form interface, contact.