求翻译：如果 a很大的话，即成为小步长的梯度下降法，由于牛顿法在误差极小点附近通常能够确定收敛更快更精确，因此算法的目的就使尽快转换为牛顿法。是什么意思？

If a very large, that is a small step gradient descent method, Newton's method because the error is usually able to determine the minimum near the convergence faster, more accurate algorithm is therefore the purpose of Newton's method to convert as soon as possible.

If a small step, that is, to become a very big drop in gradient law, as a result of Newton's law is usually near the margin of error is very small to be able to determine astringent for faster and more accurate, and therefore of the algorithm on purpose to convert to Newton's law as soon as possible

If a very big, namely becomes the small length of stride the gradient drop law, because the Newton law nearby the erroneous minimum point usually can determine restraining is quickly more precise, therefore the algorithm goal causes to transform as soon as possible into the Newton law.

If a very large, shall become a small gradient descent method for long, due to Newton's method in error can usually determine the convergence near minimizers faster and more accurate, algorithm designed to convert Newton method as soon as possible.

If a very big, namely become the gradient descent method with long little step, because the law of Newton can usually confirm that to disappear is fast more accurate very much in the error near the snack, so purpose of algorithm make as soon as possible changing Newton law into.