Advances and Analysis in Convolutional Neural Networks: A Comparative Study of AlexNet and ResNet

Jiawei Chen

2024

Abstract

Deep learning, particularly through Convolutional Neural Networks (CNNs), has significantly impacted various fields and is integral to many aspects of daily life. This study focuses on CNNs, with a specific emphasis on two foundational models: ImageNet Classification with Deep Convolutional Neural Networks (AlexNet) and Residual Network (ResNet). CNNs excel in image processing and recognition but require substantial data for effective training. AlexNet, a pioneer in the deep learning revolution, employs Rectified Linear Unit (ReLU) activation functions and Dropout techniques to mitigate overfitting but struggles with irregular and large datasets. On the other hand, ResNet introduces residual connections to address the vanishing gradient problem, although it still faces challenges related to overfitting. The paper provides a detailed comparison of CNN principles with traditional Neural Networks (NNs), highlighting the strengths and weaknesses of AlexNet and ResNet. It also explores the current challenges in deep learning, outlining potential areas for future research and development. This study offers insights into the evolution of CNN technologies and suggests directions for overcoming existing limitations and enhancing future advancements.

Download


Paper Citation


in Harvard Style

Chen J. (2024). Advances and Analysis in Convolutional Neural Networks: A Comparative Study of AlexNet and ResNet. In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML; ISBN 978-989-758-754-2, SciTePress, pages 317-321. DOI: 10.5220/0013516400004619


in Bibtex Style

@conference{daml24,
author={Jiawei Chen},
title={Advances and Analysis in Convolutional Neural Networks: A Comparative Study of AlexNet and ResNet},
booktitle={Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML},
year={2024},
pages={317-321},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013516400004619},
isbn={978-989-758-754-2},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML
TI - Advances and Analysis in Convolutional Neural Networks: A Comparative Study of AlexNet and ResNet
SN - 978-989-758-754-2
AU - Chen J.
PY - 2024
SP - 317
EP - 321
DO - 10.5220/0013516400004619
PB - SciTePress