Kernels for the ordered tree. My goal is to use the kernel for the relaitonal learning task, maybe classificaiton at the first step.
Related work:

     ILP : the traditional, and the the main stream tricks on the relational learning. I think it is just a searching method but the matching and Prolog or Lisp mechanics are powerful, so with covering algorithms and unique knowledge representation, it can solve a lot problems.
But I still dislike it, even I will learn it in the near future, because it just like a drudgery counting every thing out , :) .

    Kernels for the structured data: many thanks for the reply of  Gartner, a cool German PhD.
    Edit distance:  i.e. Levenshitein distance , I want to adopt it to the tree with the nodes for the attribute which can represent the relaitonal instances. Mr.Bohnebeck gave a similarity study of the second structure of the mRNA molecule, who also give me a lot help by email.
    LV-SVM:  the implemantation of the SVM coded by Taiwanese Chih-Jen Lin etc. Very useful! Especially the MATLAB interface.
I really appreciate the hard-working and stern German 's help.
I am working on several dataset to testify my little original idea, and I wish I can finish my paper before 10.1 (the deadline)

If I can complete the work before 9.8, I will try to send it to some conference for publication.

没什么，只是觉得还是有必要这样有气无气的让space活着，虽然不好，但是也是一种存在。
天气终于凉快了，夏天也就这样了吧。
时间在流，拉不住，我居然有这种想法，呵呵。。。。。。