LATIDIA · Robótica

Redes aproximadas de cuaterniones para la clasificación mejorada de imágenes y la detección de objetos orientados.

arXiv:2509.05512v2 Anuncio Tipo: replace-cross Resumen: Este artículo presenta Quaternion Approximate Networks (QUAN), un nuevo marco de aprendizaje profundo que aprovecha el álgebra de cuaterniones para la clasificación de imágenes equivariantes de rotación.

La noticia

arXiv:2509.05512v2 Announce Type: replace-cross Abstract: This paper introduces Quaternion Approximate Networks (QUAN), a novel deep learning framework that leverages quaternion algebra for rotation equivariant image classification and object detection. Unlike conventional quaternion neural networks attempting to operate entirely in the quaternion domain, QUAN approximates quaternion convolution through Hamilton product decomposition using real-valued operations. This approach preserves geometric properties while enabling efficient implementation with custom CUDA kernels. We introduce Independent Quaternion Batch Normalization (IQBN) for training stability and extend quaternion operations to spatial attention mechanisms. QUAN is evaluated on image classification (CIFAR-10/100, ImageNet), object detection (COCO, DOTA), and robotic perception tasks. In classification tasks, QUAN achieves higher accuracy with fewer parameters