When you are doing scientific research, you may encounter that the experimental results do not meet your expectations. What you need to do at this time is not to rush and deny yourself.
The following can be called a suitable closed-loop scientific research cycle :
- Obtain the experimental results, BAD > independent Debug, repeated inspection and scrutiny [whether the result of each sentence of code meets expectations]
- The code is OK, scrutinize the Idea, find out the intermediate results, use extreme data or settings, and prove the reasons for the badness. For example, if you have a data set with a size of 200 at this time, the analysis matrix may be 200×200. At this time, you can set some simple data sets independently, depending on the actual situation.
- Propose a solution to the bad cause, and verify it. This step is quite challenging. If you have been confined to a strange circle, you might as well jump out of your current experimental thinking, find relevant literature in a targeted manner, and see how similar problems are solved. Maybe your idea is an idea that you currently think is brand new. At this time, you may have to be careful, whether this idea has been proposed before, but the experimental results have not been very good; if it is in line with this situation, you need to go back to The nature of the problem, read more literature about the nature of the problem in your direction.
Today's sharing is here, let's work hard together~
The above process is full of tears if I talk too much, special thanks to Mr. Zhang!