Researchers have demonstrated a switchable backdoor attack that injects dual tokens through the layers of Vision Transformers, achieving up to...

A new systematic review of 55 studies traces a decade of evolution in AI-based surface defect detection, from CNNs to...

Researchers have developed S²DNet, a dynamic spectral–structural framework that adapts feature extraction and fusion to each scene, improving remote sensing...

A systematic review finds that deep learning outperforms other technologies in detecting and predicting human–wildlife conflict, with elephants, tigers, monkeys...

Researchers in Beijing built a 27-sensor deep learning system that predicts and reconstructs three-dimensional temperature and humidity fields inside solar...

A new study shows that automated metaheuristic hyper-parameter optimization substantially boosts the accuracy of Hinton's memory-efficient Forward-Forward training algorithm across...

A new generative adversarial network called SR2-GAN significantly improves multi-step radar rainfall nowcasting, cutting prediction error by up to 44.4...

A new perspective in the Journal of Medical Systems argues that clinical AI must be continuously monitored, updated, and revalidated...

Researchers have developed CausalCACTUS, a framework that generates counterfactual explanations for black-box machine learning models which simultaneously respect learned causal...

Researchers have introduced PLS-ZO, a zeroth-order optimization framework that confines gradient-free perturbations to a low-dimensional latent subspace, dramatically reducing the...