: There is research into design-history reconstruction in machined shapes using deep reinforcement learning to decompose solid models into machining features.
: Another paper, Pruning Convolutional Neural Networks with Limited Data , uses "reborn filters" to reduce information loss when compressing networks. Instead of deleting redundant channels, it develops new compact filters from existing ones to preserve performance even with minimal training data. Machined Reborn
: A research paper titled Reborn Mechanism: Rethinking the Negative Phase Information Flow in Convolutional Neural Network proposes a new activation mechanism. Unlike standard ReLU, which cuts off negative values, this mechanism "reborns" and reconstructs dead neurons to better utilize data information while keeping parameters low. : There is research into design-history reconstruction in