Story

The paper proposes a novel deep learning architecture called Convolutional Neural Fabrics (CNF) that can dynamically adapt its structure to learn effective representations for a variety of tasks, outperforming traditional convolutional neural networks on several benchmark datasets.