An Efficient Neural Architecture Search Algorithm for AutoEncoder Optimization - A Systematic Literature Review
Year: 2025
Type: review
Study design: Systematic Review
Abstract: Autoencoders have developed into neural search networks in recent years, and the majority of machine learning (ML) methods rely on the input properties to produce high-quality models. Increased dimensionality is a challenge that arises with larger datasets, which significantly reduces machine learning efficiency. In order to reduce the significant amount of high-dimensional data, researchers have developed feature reduction and selection techniques like Principal Component Analysis (PCA) and autoencoders with their different variations, including Convolutional autoencoders, Sparse autoencoders, Denoising autoencoders, Contractive autoencoders, Variational autoencoders, and Deep autoencoders. This paper discusses neural search structures, evolutionary methods, and various autoencoder types. A number of optimization techniques to increase training accuracy and decrease training time are also covered in the review. By thoroughly reviewing the literature from five scholarly databases on the subject, extracting pertinent information using a PRISMA flow diagram, and applying the proper eligibility criteria to synthesize and evaluate the best papers found, the research project adopted a thorough, detailed review. According to the research's findings, several neural search architectures and autoencoders used for data reconstruction have improved, including Venkataraman's Convolutional AEs, Baier et al.'s Self-Supervised Siamese Autoencoders, Lazebnik & Simon-Keren's Knowledge-integrated Autoencoder Model, Chen et al.'s Dirichlet Neural Architecture Search, and Zhang et al.'s Graph Hypernetworks for Neural Architecture Search.
Source: UMYU Scientifica
DOI: 10.56919/2543.002
Authors: Samuel Michael Ogbe, Ahmed Aliyu Abubakar
Institution: Kaduna State University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4415823699