Peng Li, Lu Huang, Jin Li, Haiyan Yan, Dongjing Shan. "Graph-based vision transformer with sparsity for training on small datasets from scratch." Scientific Reports, 2025. doi:10.1038/s41598-025-10408-0