This book explores the emerging role of performance predictors in evolutionary neural architecture search (ENAS), highlighting representative methods and practical applications that make this field both timely and impactful. By bridging performance prediction with evolutionary optimization, it addresses one of the most pressing challenges in deep learning: how to efficiently design and evaluate neural architectures without incurring prohibitive computational costs.

The book provides a systematic overview of predictor-driven approaches across diverse neural network model families, including graph neural networks, convolutional neural networks, and fuzzy neural networks. It introduces accuracy predictors as well as rank-aware predictors, illustrating how these methods enhance the efficiency, scalability, and generalizability of neural architecture search. In addition to results on widely used benchmark datasets, the book emphasizes practical applications such as defect detection and medical image segmentation, showcasing how predictor-guided ENAS delivers both research insights and real-world impact.

By engaging with this book, readers will gain a clear understanding of how performance predictors accelerate ENAS, discover both classical techniques and recent advances, and appreciate the methodological and applied value of predictor-guided architecture design. The book equips its audience with frameworks to evaluate and extend predictor-based methods, positioning them at the intersection of evolutionary computation, performance prediction, and neural architecture search. Additionally, the code related to the book will be available as open source.

This volume is intended for researchers, graduate students, and professionals seeking to deepen their expertise in evolutionary computation, neural networks, and neural architecture search. A foundational background in these areas will facilitate full engagement with the material and enable readers to leverage the presented concepts for both academic inquiry and applied innovation.

Focuses on performance predictor-driven approaches in evolutionary neural architecture search Bridges methodological advances with real-world applications such as defect detection under uncertainty Provides a rigorous framework combining evolutionary computation, predictive modeling, and deep learning

Autorentext

Nan Li is a Lecturer in Institute of Big Data Science and Industry, Shanxi University, China. His research interests include computational intelligence and machine learning. Li Nan has been selected for the inaugural China Association for Science and Technology (CAST) Young Talent Support Program's Doctoral Student Special Project. To date, he has published 20 papers in prestigious journals and conferences including ACM CSUR, NeurIPS, IJCAI, IEEE TEVC, IEEE TFS, and IEEE TCYB, including 3 ESI highly cited papers, 1 hot paper, and 3 research frontier papers, garnering over 400 Google citations. He serves as a reviewer for over 30 SCI/EI journals, including 20 Q1 journals of the Chinese Academy of Sciences and 3 international conferences as PC/TPC reviewer. 

Lianbo Ma is a Professor at Northeastern University, China.. His current research interests include computational intelligence and machine learning. He has published over 100 papers in prestigious journals and top conferences including IEEE/ACM TON, IEEE TMC, IEEE TEVC, IEEE TCYB, IEEE TC, IEEE TNNLS, IEEE TCDS, IEEE COMST, ICML, ACM MM, NeurIPS, IJCAI, AAAI, and other authoritative journals and top conferences. Representative works have been selected as ESI Highly Cited Papers and ESI Hot Papers, and received the IEEE Computer Society Outstanding Paper Award.

Yuhua Qian is a Professor and Director of Institute of Big Data Science and Industry, Shanxi University, China. He is also the Vice-President (Research) of Shanxi University. He is actively pursuing research in pattern recognition, feature selection, rough set theory, granular computing, and artificial intelligence. He has published more than 200 articles on these topics in international journals. He has published in AIJ, IEEE TPAMI, JMLR, ML, ACM TIS, ACM TKDD, IEEE TEVC, IEEE TNNLS, IEEE TKDE, IEEE TFS, IEEE TSMC. He holds four invention patents. These research outcomes have been widely adopted by scholars globally in fields including remote sensing image analysis, medical diagnostic data analysis, biological data mining and social network analysis.

Bing Xue is a Professor of Artificial Intelligence at Victoria University of Wellington, New Zealand, a Fellow of IEEE, and a Fellow of Engineering New Zealand. Her research focuses on evolutionary computation, symbolic regression, feature selection, multi-objective optimisation, and automated machine learning. She has been awarded two Marsden grants, Large-scale Evolutionary Feature Selection for Classification in 2016 and Evolutionary Automated Design of Deep Convolutional Neural Networks for Image Classification in 2019. The latter project aimed to improve the design of deep convolutional neural networks by using evolutionary computation-based approaches. Deep convolutional neural networks are used to interpret images in many situations, such as security, self-driving vehicles, and medicine. Improvements to their design would lead to better "classification accuracy, improved speed, simplicity, and interpretability”. 

Mengjie Zhang is a Professor of Computer Science at Victoria University of Wellington, New Zealand. He is a Fellow of the Royal Society of New Zealand, a Fellow of IEEE, and a Fellow of Engineering New Zealand. He receives the “EvoStar/SPECIES Award for Outstanding Contribution to Evolutionary Computation in Europe” in 2023, the Australasian Artificial Intelligence Distinguished Research Contribution Award in 2024, and the ACM SIGEVO Outstanding Contribution Award in 2025.  Since 2007, he has been listed as a top five (currently No. 2) world genetic programming researchers by the GP bibliography.  He is also a Clarivate Highly Cited Researcher in Computer Science in 2023 and 2024.

Inhalt

"I-Fundamentals and Background".- "1.Evolutionary Neural Architecture Search".- "Efficient Performance Evaluation".- "Performance Predictor".- "II-Performance Predictor-Based ENAS Methods".- "Compact Encoding Predictor for Automated GNN Design".- "5.Listwise Ranking Predictor for Automated CNN Design".- "A Pareto-wise Ranking Classifier for Automated CNN Design".- "III-Advanced Performance Predictors".- "Transferable Relativistic Predictor".- "Single-domain Generalized Predictor".- "IV Performance Predictor-Based ENAS Applications".- "FNN Architecture Search for Defect Recognition under Uncertainty".- "Automated FNN Design for Defect Detection via Multiobjective
Optimization".- "Rank Predictor-assisted Architecture Design for Biomedical Image Segmentation".- "V.Conclusions and Future Research Direction".- "12.Conclusions and Future Work".