Randomness and Interpolation Improve Gradient Descent

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Jiawen, Lefevre, Pascal, Majeed, Anwar Pp Abdul
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914093723549696
author Li, Jiawen
Lefevre, Pascal
Majeed, Anwar Pp Abdul
author_facet Li, Jiawen
Lefevre, Pascal
Majeed, Anwar Pp Abdul
contents Based on Stochastic Gradient Descent (SGD), the paper introduces two optimizers, named Interpolational Accelerating Gradient Descent (IAGD) as well as Noise-Regularized Stochastic Gradient Descent (NRSGD). IAGD leverages second-order Newton Interpolation to expedite the convergence process during training, assuming relevancy in gradients between iterations. To avoid over-fitting, NRSGD incorporates a noise regularization technique that introduces controlled noise to the gradients during the optimization process. Comparative experiments of this research are conducted on the CIFAR-10, and CIFAR-100 datasets, benchmarking different CNNs(Convolutional Neural Networks) with IAGD and NRSGD against classical optimizers in Keras Package. Results demonstrate the potential of those two viable improvement methods in SGD, implicating the effectiveness of the advancements.
format Preprint
id arxiv_https___arxiv_org_abs_2510_13040
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Randomness and Interpolation Improve Gradient Descent
Li, Jiawen
Lefevre, Pascal
Majeed, Anwar Pp Abdul
Machine Learning
Artificial Intelligence
Based on Stochastic Gradient Descent (SGD), the paper introduces two optimizers, named Interpolational Accelerating Gradient Descent (IAGD) as well as Noise-Regularized Stochastic Gradient Descent (NRSGD). IAGD leverages second-order Newton Interpolation to expedite the convergence process during training, assuming relevancy in gradients between iterations. To avoid over-fitting, NRSGD incorporates a noise regularization technique that introduces controlled noise to the gradients during the optimization process. Comparative experiments of this research are conducted on the CIFAR-10, and CIFAR-100 datasets, benchmarking different CNNs(Convolutional Neural Networks) with IAGD and NRSGD against classical optimizers in Keras Package. Results demonstrate the potential of those two viable improvement methods in SGD, implicating the effectiveness of the advancements.
title Randomness and Interpolation Improve Gradient Descent
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2510.13040