Original scientific paper
https://doi.org/10.24138/jcomss-2023-0124
Spectral Proximal Method and Saliency Matrix for Robust Deep Learning Optimization
Cherng-Liin Yong
orcid.org/0000-0003-0952-5252
; Universiti Tunku Abdul Rahman, Jalan Sungai Long, Bandar Sungai Long, Kajang, Selangor, Malaysia
*
Ban-Hoe Kwan
; Universiti Tunku Abdul Rahman, Jalan Sungai Long, Bandar Sungai Long, Kajang, Selangor, Malaysia
Danny-Wee-Kiat Ng
orcid.org/0000-0001-9972-2676
Hong-Seng Sim
; Universiti Tunku Abdul Rahman, Jalan Sungai Long, Bandar Sungai Long, Kajang, Selangor, Malaysia
* Corresponding author.
Abstract
This paper presents a new training optimizer for
deep learning models, called the Spectral Proximal (SP) method
with saliency matrix, that aims to improve their ability to
generalize to new data. Generalization is the measure of how well
a model can perform on data that it has not seen during training.
The SP method addresses a pair of hurdles affecting generalization: the problem of gradient confusion within complex model
architectures and the limited availability of training data. The
key innovation of the SP method is the use of a proximal operator
with a saliency matrix, which adjusts the descent direction based
on the importance of each parameter and avoids overfit issues.
This leads to improved performance on image classification
(MNIST and CIFAR-10) and object detection (YOLOv7) tasks
and better ability to generalize to new data. We conducted a
comprehensive inquiry by performing experiments on various
configurations while controlling for potential confounding factors.
The SP method consistently outperformed the baseline method
based on the results.
Keywords
spectral proximal method; saliency matrix; deep learning; machine vision; optimization algorithm
Hrčak ID:
315205
URI
Publication date:
26.2.2024.
Visits: 540 *