Abstract:
"his study presents the development of a novel algorithm, QAPNet(Quantization Attention Pruning Network), and its comprehensive evaluation, demonstrating its superior performance compared to traditional deep learning models
such as VGG (Visual Geometry Group) and ResNet (Residual Network) for skin
cancer prediction. QAPNet achieves an impressive accuracy of 99.47% and an
F1-score of 99.5%, showcasing exceptional precision and robustness. In contrast,
VGG reaches 97.78% accuracy with a 97.92% F1-score, and ResNet achieves
96.67% accuracy and a 96.91% F1-score. The analysis of training efficiency indicates that QAPNet converges more rapidly due to its advanced architecture and
optimized hyperparameters, including a lower learning rate and smaller batch
size. The effectiveness of the Adam optimizer is highlighted through sensitivity
analysis, further enhancing QAPNet's performance. Moreover, QAPNet demonstrates superior efficiency with fewer parameters and lower floating-point operations (FLOPs) compared to VGG and ResNet. This reduction in error rates
reinforces QAPNet's advantage over traditional models. Overall, QAPNet's combination of high accuracy, efficiency, and robustness positions it as a compelling
choice for applications requiring both precision and real-time processing capabilities. While VGG and ResNet are effective within their domains, they exhibit
limitations that QAPNet overcomes through its innovative design. Future research should focus on refining these models and exploring their performance in
diverse practical scenarios to fully leverage their capabilities and potential for
advanced applications."
