AN EXTENSIVE EXAMINATION OF PROSPECTS AND OBSTACLES IN ENSEMBLE DEEP LEARNING

Abstract

: In the realm of machine learning, two dominant approaches surpass traditional algorithms: ensemble learning and deep learning. Ensemble learning integrates multiple base models within a single framework to yield a more robust model that surpasses individual ones. The effectiveness of an ensemble method hinges on factors like the training of base models and their fusion. Common approaches to constructing successful ensemble models are well-documented across various domains. Conversely, deep learning models have significantly enhanced predictive accuracy across diverse domains, leveraging their diverse architectures and automated feature extraction capabilities. However, the main challenge with deep learning lies in the intricate tuning of hyper-parameters, demanding substantial expertise and time. Recent research endeavors have explored integrating ensemble learning with deep learning to tackle this challenge. While most efforts focus on basic ensemble methods, this review delves into a comprehensive examination of strategies for ensemble learning, particularly within the context of deep learning. It meticulously discusses the factors influencing ensemble method success and categorizes numerous research endeavors that have applied ensemble learning across a broad spectrum of domains.