An impressing seminar, about Machine Learning

Machine learning is not a new method, but Doctor Chen Yu Zong's （http://bidd.nus.edu.sg/group/bidd.htm） lecture really impressed me. I have some thoughts, mostly about science research. Although some of them have been in my mind for a long time, they have never been considered in such a practical way.

Fisrt, Being Scrupulous. Details may be very important, not in the experiments but also in the data analysis. For instance, he said, choosing suitable training samples can be the most important thing in machine learning. Methods, such as SVM, are tumblers. You can always find a solution from a group of samples no matter how you do the sampling. Sometimes the results may seem perfect, but indeed, there are several flaws. What I want to say is: some methods seem perfect, some results seem perfect. However, we should not be satisfied by them, while they could be mistakes.

Secondly, to be an expert in the field before action. Doctor Zong did a thorough statistic about previous studies on cancer gene biomarker. In this way, his research is based on a good investigation. He knows what is the problem in this kind of study. In contrast, many researchers just read few papers before their proposal. Then what can they do with so little relevant knowledge? A couple of days ago, Prof. Shen said: ”We are in the best institute of China, then don’t you feel awful if you are doing things which can be finished by a regular researcher? ” We must do better, upon the knowledge of previous scientists, that’s how a researcher looks like!

This blog hasn’t been updated for a long time. I’ll try my best to keep this site active. Thank you for your visit.