(1)
Bu, F.; Chang, D. E. Feedback Gradient Descent: Efficient and Stable Optimization With Orthogonality for DNNs. AAAI 2022, 36, 6106-6114.