Evolutionary neural architecture search: a survey

Memetic differential evolution with niching strategy for efficient CNN architecture search

Training to Design Convolutional Neural Networks

It is demonstrated that network architecture and its coeficients can be learned together by unifying concepts of evolutionary search within a population based traditional training process.

Evolutionary Machine Learning Meets Self-Supervised Learning: A Comprehensive Survey

Research that combines Evolutionary Machine Learning and Self-Supervised Learning has been steadily increasing in recent years, suggesting that the combination of these two areas can help both in shaping evolutionary processes and in automating the design of deep neural networks, while also reducing the need for labelled data. Yet, no survey details how these two areas are used together. To help with this, we introduce Evolutionary Self-Supervised Learning as a research area and propose a taxonomy based on two core directions, evolution applied to Self-Supervised learning and Self-Supervised learning applied to evolution. For each direction, we categorise and discuss different approaches, compare the performance of different works, and discuss observed trends and their impact on performance and computational cost. Following this, we identify six open challenges around pretext task design, ablation studies with label scarcity, and the design of reliable and low-cost fitness metrics that take into account self-supervised mechanics. Based on these challenges, we propose research directions for researchers within the field.

Optimization of Artificial IntelligenceDriven Evolutionary Algorithms for Complex Engineering Problems

Currently, technologies such as convolutional neural network (CNNs) and deep Q network (DQNs) are undergoing intensive research and rapid development, driving vigorous advancement in the field of artificial intelligence. Nevertheless, there is still room for improvement in addressing practical industrial problems and enhancing learning efficiency and accuracy. To address the core challenges of traditional evolutionary algorithms (EAs) in complex optimization problems, such as insufficient scalability, limited environmental adaptability, and low computational efficiency, this paper proposes an evolutionary algorithm optimization framework (DLRL-EAF) that fuses deep learning and reinforcement learning. To verify the effectiveness of the proposed method, six standard test functions (Sphere, Rastrigin, Griewank, etc.) and three practical engineering optimization problems (mechanical parts design, logistics path planning, photovoltaic array layout) are selected for comparison experiments. The performance of DLRL-EAF is evaluated using the standard genetic algorithm (SGA), the particle swarm optimization algorithm (PSO), and the adaptive genetic algorithm (AGA). Experimental results show that DLRLEAF improves the accuracy of optimal solutions by an average of 23.6%, accelerates iterative convergence by 31.2%, demonstrates greater stability in high-dimensional, complex problems, and improves scalability by more than 40%. At the same time, the proposed method significantly reduces the time and resource costs of problem-solving in practical engineering applications and demonstrates its practical value in industrial settings.OPEN ACCESS Received: 27/12/2025 Accepted: 26/03/2026

Evolutionary Training of Neural Networks: The Role of Crossover Operators in Genetic Algorithms Compared with Backpropagation

Training neural networks with gradient-based methods such as backpropagation is the dominant paradigm, but it depends on differentiable loss functions and is sensitive to initialization and local minima. Evolutionary algorithms offer a gradient-free alternative, yet the influence of their internal operators on training quality remains insufficiently characterized. This study presents a systematic comparison of backpropagation and ten variants of a genetic algorithm (GA) for training multi-layer perceptrons (MLPs), with particular focus on the role of crossover operators. The evaluation covers four MLP architectures and ten classification datasets from the UCI Machine Learning Repository, differing in sample size, dimensionality, and number of classes. Each configuration was assessed using stratified 4-fold cross-validation with 30 independent repetitions, and accuracy served as the primary performance metric, with macro-F1 reported to assess classifier behavior on class-imbalanced datasets. Backpropagation achieved higher mean accuracy than every GA variant on nine of the ten datasets, with the largest margins on high-dimensional problems. The genetic algorithm proved competitive on simpler, class-balanced datasets, where its better-performing variants matched the gradient-based baseline within one to two percentage points, and, on the Heart disease dataset, every GA variant reached a higher mean accuracy than backpropagation across all four architectures, though absolute performance remained modest on this five-class problem. Among crossover operators, BLX-α and BLX-α-β combined with tournament selection and a high crossover probability yielded the strongest configurations, while averaging crossover performed worst, as it restricts offspring to the midpoint of the parents and cannot explore beyond the range already present in the population. Tournament selection consistently led to higher mean accuracy than roulette-wheel selection, and shallow but moderately wide architectures, which encode fewer trainable parameters and thus a shorter chromosome, proved more amenable to evolutionary training than the two-layer alternative. These findings clarify when gradient-free training is competitive and which evolutionary operators drive its effectiveness.

RLDMGO: a reinforcement learning-driven hybrid framework for complex optimization problems