[2603.20920] Democratizing AI: A Comparative Study in Deep Learning Efficiency and Future Trends in Computational Processing

Computer Science > Performance arXiv:2603.20920 (cs) [Submitted on 21 Mar 2026] Title:Democratizing AI: A Comparative Study in Deep Learning Efficiency and Future Trends in Computational Processing Authors:Lisan Al Amin, Md Ismail Hossain, Rupak Kumar Das, Mahbubul Islam, Saddam Mukta, Abdulaziz Tabbakh View a PDF of the paper titled Democratizing AI: A Comparative Study in Deep Learning Efficiency and Future Trends in Computational Processing, by Lisan Al Amin and 5 other authors View PDF HTML (experimental) Abstract:The exponential growth in data has intensified the demand for computational power to train large-scale deep learning models. However, the rapid growth in model size and complexity raises concerns about equal and fair access to computational resources, particularly under increasing energy and infrastructure constraints. GPUs have emerged as essential for accelerating such workloads. This study benchmarks four deep learning models (Conv6, VGG16, ResNet18, CycleGAN) using TensorFlow and PyTorch on Intel Xeon CPUs and NVIDIA Tesla T4 GPUs. Our experiments demonstrate that, on average, GPU training achieves speedups ranging from 11x to 246x depending on model complexity, with lightweight models (Conv6) showing the highest acceleration (246x), mid-sized models (VGG16, ResNet18) achieving 51-116x speedups, and complex generative models (CycleGAN) reaching 11x improvements compared to CPU training. Additionally, in our PyTorch vs. TensorFlow comparison, we observed t...