SHOT: Suppressing the Hessian along the Optimization Trajectory for Gradient-Based Meta-Learning

To validate the effectiveness of SHOT, we conduct empirical tests on standard few-shot learning tasks and qualitatively analyze its dynamics.