Superior efficacy of trust region method in improving generalization errors for regression tasks in scientific machine learning.
Scientific machine learning training problems benefit from second-order solvers for regression tasks.