This is the first time a postgraduate tutor has given us speech and guidance, and it is probably the only time in the first semester. I tried my best to grasp this very hard-won opportunity and didn't easily let go of every word the instructor had said. Then take advantage of the day when the memory has not faded, carefully extract it for future reference.
1. Study paper
Search ability is a very important ability, and access to the Internet is the most basic requirement for a computer major. Don't rely on the papers recommended by your supervisor, you need to learn how to search for the resources you want.
Need to do it during graduate school, follow the top conferences! ! !
Every month, you have to look at the list that will be updated at the top, otherwise you will have no confidence to speak! !
Be sure to attach great importance to articles from top conferences! ! ! ! !
Journal papers are generally the results of one to two years.
Try not to make improvements to journal articles. Timeliness will be questioned!
Conference papers are generally the results within six months! ! !
Papers from top conferences represent the forefront of research.
Most of the articles published in top conferences will open the source code and open the data set, which is very convenient for latecomers to reproduce.
Improve the papers of top conferences. Timeliness and paper quality are guaranteed, and papers will not be questioned! ! !
Remember, if you want the best and top people to learn and improve their papers, the quality will be higher.
The most effective way to improve: reproduce the code of the top meeting.
2. Thesis direction
The innovative points of writing a paper are mainly divided into the following three points:
1. Citation-based innovation Cite the
core technology of the paper in another context to this context.
For example: I
learned a paper "based on cloth defect detection" and learned the core algorithm.
Then use the core algorithm to a field I set "based on steel defect detection", and make some adjustments to the algorithm.
But requirements: the
cited papers must have certain aspects superior to other papers in the same field.
2. Integrated innovation
The application of new knowledge and technology in another field to this field.
For example:
interdisciplinary, combining the field of biology and deep learning to study a comprehensive problem.
For example, through data collection of human joint points and apes and monkey joint points, computer analysis of human posture, combined with biological technology, to study the related issues of human evolution.
3. Improved innovation The
most extensive one.
The most recommended mode is to improve the latest top conference papers.
Standing on the shoulders of giants.
3. Write a paper
Changes in the postgraduate study stage
The first stage: a
wide range, a broad understanding of the various branches of this direction, and understanding of its most advanced development status, etc.
The second stage: the
scope changes from large to small, gradually choosing several points of interest, narrowing the research field, and focusing on one or several branches.
The third stage: the
scope changes from small to large. After determining your research field and direction, search for relevant papers from the latest top conferences. Read a lot of papers in the field.
The fourth stage
has large and small scope, focusing on this direction, the best individuals and laboratories in this field. Open their homepage and learn about their series of papers. Even, predict their next research direction.
For the preliminary work, you must read a lot of papers from top conferences and have enough accumulation. After reading each paper, learn to summarize and generalize the revised paper, and then you can add the citation of the paper to the Introduction of the paper.
In a nutshell, as long as the work of reading a large number of papers in the early stage is in place, the writing of the first and second parts of the paper will be straightforward, and a large number of articles can be cited in one sentence.
Preliminary work: Actively reproduce the code of other people's papers. On the one hand, you need to learn, improve, and innovate the paper code. On the other hand, in the experimental analysis part of the paper, you need to compare the experiment with the original code and run the code.
If the paper is accepted by an international conference, the laboratory will be reimbursed for the full cost of the road. It is necessary to prepare English ppt, prepare English speech, and prepare English question and answer. Usually there will be half a year to prepare from the appointment to the meeting.
The basic requirements of reporting PPT:
1. A lot of pictures, less text
2. The emphasis
4. Essay writing
The ability to write papers is very important! ! !
1. Pay attention to logical thinking;
2. The content of the paper should not be spoken, the subject, predicate and object are complete;
3. The writing ability is a bonus item of the paper. In addition to reading the content of the paper, the reviewer will also pay attention to the writing ability
4. You can consider writing a Chinese paper first , Complete subject, predicate and object, standard writing. Then use a translation tool to translate it into English for inspection.
Summary:
Standard and rigorous writing is the facade of the thesis.
5. Academic bottom line
Some key points about plagiarism:
1. Copy a sentence from someone else's paper, even if it is Introduction, and add a citation to that paper. This behavior is still plagiarism! ! !
Correct approach: Use your own language to summarize the content of other people's papers, and then add citations to the paper.
2. I copied my previous paper, that is, one or two sentences of this paper are copied from my previous paper. This behavior is still plagiarism! ! ! Self-plagiarism! ! !
The correct approach: Even if these two papers are of the same background, the research is progressive. A new paper Introduction must also be rewritten.
3. Take the results of others and write in your own words.
There is no doubt that plagiarism! ! !
About fraud:
Do not falsify data, which requires us to run the model and run the code by ourselves. The data must be run out of the code, not artificial.
Finally, at the graduate level, no matter what you do well or not, you must do your own work.
If you don't do it well this time, it doesn't mean you will continue to do it badly in the future.
But if you touch the academic bottom line, there will be no future opportunities.