麻烦大家帮我看看，这个教授的回复，我应该怎么回信呢，谢谢

Thank you for forwarding this interesting work.  My group has some experience with the sensitivities of stochastic gradient methods to learning rates, but not with the second-order methods you pursue here.  We ended up developing a non-stochastic online alternative that avoids learning rates entirely: 论文名
Your comment about "reusing the previous information" sounds related.

I recently returned from a very-exciting NIPS workshop on variational inference, which had many talks and posters relevant to your interests:
  workshop 网址

我应该看教授的论文，然后写review吗？

为什么不呢?