Multi Objective Neural Architecture Search

Multi-objective neural architecture search (MONAS) automates the design of neural networks by optimizing multiple objectives simultaneously, such as accuracy, model size, and inference speed. Current research focuses on developing efficient algorithms, often employing evolutionary methods or Bayesian optimization, to explore the vast search space and discover Pareto-optimal architectures—models that offer the best trade-offs between competing objectives. This field is significant because it addresses the limitations of traditional single-objective NAS by generating models better suited for real-world deployment on resource-constrained devices, improving both performance and efficiency across diverse applications.