Time Trials on Second-Order and Variable-Learning-Rate Algorithms