Parametric NAS: Revolutionizing Neural Architecture Search for Lung Colon Cancer Classification

Neural Architecture Search (NAS), called Parametric NAS, to address the challenges in automatically designing efficient neural network architectures for Colon Cancer classification tasks. By incorporating techniques such as Sequential Model-Based Optimization and Pareto optimality, Parametric NAS integrates multiple optimizations into a unified approach, thereby simplifying the complexity of the search space. This methodology offers a more effective and efficient solution than previous ones by significantly reducing the complexity of architecture and weight optimization. The proposed Parametric NAS introduces a novel search space and develops two networks within this framework. Furthermore, the paper presents architecture saliency, a novel selection criterion for choosing optimal NAS architectures based on the squared change in network loss. referred to as architecture saliency. The paper concludes by demonstrating the efficacy of the Parametric NAS approach through experiments, showcasing an ensemble of ResNet50 and Elastic Net for pixel and image-level classification in identifying lung and colon cancer cells. This ensemble model is a significant development in the neural architecture search for Lung and Colon Cancer classification, which provides better accuracy and efficiency than previous approaches.