Tools
Tools
-, Gaurav (2024) A Deep Learning Feature-Ranked Backpropagation Framework for Sustainability. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (3MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (792kB) | Preview |
Abstract
Backpropagation method is used for training neural networks, it adjusts weights through error feedback from layer wise predictions, improving the accuracy of the model over epochs. Feature-ranked backpropagation prioritizes features during training by adjusting the weights based on their influence while masking the weights with less influence on the output to improve learning efficiency. Sustainability in deep learning refers to the process of developing artificial intelligence (AI) models that reduce computational resource requirements such as reducing computational time requirements( from 1 hour to 45 minutes). This paper proposes a novel framework combining Feature-Ranked Backpropagation with established deep learning architectures such as AlexNet, ResNet-18, and VGGNet-19 to reduce the use of resources such as training time, computational requirements(RAM, GPU, and Processors), energy consumption and CO2 emissions. The proposed method dynamically prioritizes the most critical weights for updates, guided by feature-ranking techniques derived from gradient-based saliency maps. The models trained on the subset of ImageNet dataset containing 66,433 images across 50 categories for image classification. The framework achieves notable reductions up to 20% savings in energy usage and 15% in training times compared to standard training approaches, with comparable accuracy metrics. This paper lays foundation for training deep learning models efficiently and sustainably.
| Item Type: | Thesis (Masters) |
|---|---|
| Supervisors: | Name Email Stynes, Paul UNSPECIFIED |
| Uncontrolled Keywords: | Feature Ranking; Backpropagation; Deep Learning; AlexNet; ResNet; VGGNet; Image Classification; Sustainability; Training Efficiency |
| Subjects: | Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence |
| Divisions: | School of Computing > Master of Science in Artificial Intelligence |
| Depositing User: | Ciara O'Brien |
| Date Deposited: | 19 Jun 2025 14:59 |
| Last Modified: | 19 Jun 2025 14:59 |
| URI: | https://norma.ncirl.ie/id/eprint/7937 |
Actions (login required)
 |
View Item |