When it comes to thesis writing, some students may think that thesis writing is: how to structure the thesis, how to write, and about grammar. I prefer to think that thesis writing is from the moment you enter the tutor's project team , it will start when you get your subject.
Why do you say that?
Because your thesis is composed of your opening report, research content, including subsequent methodological innovations, experimental parts, etc.
Regarding the writing of thesis, I want to discuss with you in this article how to carry out the project work from the postgraduate stage, to the submission and publication of the thesis in the later stage, and some key issues, and share some experience and experience of my personal understanding. Since everyone studies different majors, I will only talk about some common issues.
First of all, let me introduce myself. I am a doctor of engineering from the Chinese Academy of Sciences. I am currently a postdoctoral fellow at the Chinese Academy of Sciences. My research directions are mainly nonlinear system modeling control methods, intelligent optimization algorithms, and image processing algorithms. I am currently in charge of the National Nature Approach Qingqing project department as the project leader, and I am currently participating in the research and development of major national space models. Regarding the guidance of doctoral students, I have published more than ten papers in domestic and foreign journals as a master student. Has three invention patents.
Regarding the difficulties and precautions in the process of carrying out scientific research, writing and publishing papers, I have summarized four aspects by myself,
The first aspect is how to find the entry point for postgraduate research;
The second aspect is how to innovate the methodology in the subject;
The third is how to select a journal in the paper submission;
The fourth is how to improve our own thesis writing level.
For the students who are currently in the postgraduate stage, the status they are in can be roughly divided into the following categories:
The first one may be that you have just entered the postgraduate stage and have not done research in the undergraduate stage. At this time, you will be confused. The tutor has given you a topic and you don’t know how to carry it out. I can't give myself some advice.
The second type of students may have been neglected in the past two years, or were dragged away by the teacher to do odd jobs. Now that they are about to graduate, the pressure is relatively high. At this time, they are faced with such problems as how to do their own projects, how to optimize and innovate. .
When it comes to writing a dissertation, everyone will encounter such a problem.
In addition, for the first type of students, in fact, for postgraduate students, you are still in class in the first year and start research in the second year. point, then you may soon become the second class of students. You will find that these two years have passed in the blink of an eye, and you will find that the time is not enough when you are really facing graduation.
Therefore, among the four aspects I just summarized, the first aspect is how to find the entry point for the postgraduate course stage is very important. If you spend too long at this stage, then your body will naturally be compressed, and if you practice with confidence later, you may face delays.
1. How to find the entry point for postgraduate research?
Today's first topic, I want to talk to you about this issue, that is, how do I find an entry point for a topic?
1.1 Pioneering Research
First of all, regarding the writing of thesis, I divide the postgraduate topics into two categories. The first category I think is pioneering research, and the second category is inherited research. For example, in the research group you entered, your supervisor and the research group are engaged in intelligent robot technology. There are three key technologies in the field of intelligent robotics, one is environmental perception, the other is behavior planning, and the third is Xie Dongfengzhi. Then in this research group, your mentor and senior brothers and sisters have been engaged in collaborative control before, they only engage in collaborative control, and when you enter the research group, the instructor just wants to develop something, he will give this task to I gave it to you, so you became the first person in the project team to eat crabs. At this time, you don't have a mentor or a senior brother or sister to ask for advice. You have to do your own research and find a way. How to find the entry point? How to start the following research?
1.2 Inheritance Research
What is Inheritance Research? For example, when you enter this research group, the supervisor needs to do action control in advance. Your senior brother and sister have already done some pioneering work, but the supervisor thinks that the work done by the senior sister may have room for improvement in the performance indicators of the model. Maybe you will arrange such a topic for you to continue his work on the shoulders of your predecessors, that is, how to improve the performance of the model? How to improve the speed of model processing? It is also possible that the models of your brothers and sisters have been stuck in the imitation stage, and the instructor wants to go a step further and carry out hardware transplantation. Maybe if the original algorithm is very complex, how to make it easier, how to optimize the model of the previous brothers and sisters? Comparing the two big studies, this one is an inherited study. You are standing on the shoulders of Migration, and the brothers and sisters of the same school generally will not hide from you. If you have any questions, they will tell you some of the process of conducting research at that time. Compared with pioneering research, inheritance research saves such an entry process.
So how to do pioneering research? When many students are doing pioneering research, the tutor gives him the project, and he only really researches it when writing the opening report. You said that you spend a lot of time researching, but what is the focus of your research? It may be just to search for keywords from the knowledge network, and then read some papers to see what methods others have, what methods they use, and how innovative the methods are. On the contrary, many students ignore the review literature and the introduction part. They think that this is used by the author to make up the number of words when writing, so they ignore this part instead. But in fact, this part is often very important. It is crucial for you to find the entry point of the thesis and form a systematic understanding of the topic or direction of your current research.
What is a review means that for a certain topic, some authors have summarized and analyzed the data, materials and methodologies in a large number of papers on the current topic, and formed a new text. The review does not give you the methodology, it is for you to get started and innovate, because many students are faced with the problem of getting started, and they have only a half-knowledge about the topics assigned by the tutor, and they don't know how to start and how to carry out the following work.
I divide reviews into applied reviews and research reviews. Applied reviews are like the one on the left, and research reviews are like the research review on modeling and control of continuous flight systems that I studied. My supervisor will give you such a research field, It lets you explore. If I went to HowNet or Google Academics to investigate the review of this type of practical flight system field from the very beginning, I would use the research photos to find out what methods are currently available in this field and what is the current research level of this method at home and abroad , What kind of indicator requirements or level requirements do I want to achieve, and what level of papers I can publish a very systematic understanding.
2. How to carry out methodological innovation in the project?
What is the second summary? It is methodological. For example, after I have formed a systematic understanding of this field, I will focus on the innovation of methodology, such as algorithms or models of basic classes. For example, the one in the lower right corner is a data-driven algorithm based on evolution. I am sure that the one below is in my parameter gift box. If I want to use this type of algorithm, then I will research the literature on this type of evolutionary algorithm. What are the current algorithm models of evolutionary algorithms, what improvements have been made, and what level has been achieved so far, and in this review, whether it is a review of the application field or a breakdown of the methodology, he will discuss the current methods in this field , Its shortcomings, there will be an explanation. This short board is likely to be what you want to improve in the future. The direction of this idea will remind you, for example, what is the reason for the current algorithm, which may cause it to be ineffective in a certain aspect, or speed, processing If the speed is not good, then I will pass such a shortcoming, and when I may face him to improve, I just have a way, not like a headless chicken.
At this time, many students choose a model at random and start looking at it. What should they do at this time? For example, for the models of these four scenarios, for each targeted model, find the corresponding literature for intensive reading. You can read their principles. Once you introduce the derivation of mathematical formulas, after you read it, you find that every Individuals have different levels. Some models require mathematical derivation. You look like a bible. If you are just in the first year of this hero pass team, you may have time. I will attack it and I will After the strong attack, I may have improved my own level. The attack is not enough. Anyway, I have plenty of time behind me, but if you are already facing the pressure of graduation, you still find it very difficult. To understand the model at hand, you are likely to face the fact that you have made no progress for several months, and the last time is wasted.
Often when some authors are writing a thesis, they have some pitfalls or problems in the application process, experiment process, or code implementation process, and they will not write them in the paper. Some problems only need you to start. Reproduce the code, start to do experiments, and only when you start to implement, can you know what the application field is like. For example, for target detection, you only have to find a model. After you start using it, you will go down a whole set of methods and pass through the data set. After training, apply it to a certain type of specific graphics, target recognition, how to simulate the code, and how to run the laboratory? What was the final result? Only after doing this, can you really understand what you want to do? This is better than reading ten, 20, and 30 literatures, but instead, some students read them as soon as they enter the graduate school. They just sit at their desks and research literatures every day. The article is very emotional, but after seeing 20 articles and 30 articles, to be honest, the final effect is very poor.
Whether it is groundbreaking research or inherited research, model improvement and model replacement are the top priorities. Whether it is academic paper writing, journal papers, or your final graduation thesis, this is the top priority. If there is no model improvement, no model replacement, no method innovation, you will not be able to publish, and you will not be able to change.
On this basis, how to do model improvement and model replacement? That is how to innovate the method.
What is innovation? That is, no one has done it before, you are the first to eat crabs, the first to make it, and the effect is not bad, this is innovation.
I mainly talk about innovations in the field of engineering, electronics, or computers, control fields and other fields.
2.1 Principle innovation
How to innovate? That is, you propose a new model in this field. The first is to improve the model itself, and the second is to reference some ideas in the model. For example, how did many people innovate in the past few years? They replaced the regular convolutional structure inside with a depthwise separable convolutional structure. The depthwise separable convolutional structure is relatively faster to process, which can increase the data speed of the entire network, and can also improve the object detection speed. Many people think of ideas, such as replacing the conventional and established structure inside. This is a kind of innovation, because you have proposed a new model.
The second is to introduce an idea and a mechanism into the existing model, many of which do pose estimation or target detection, and introduce a channel into it. The attention mechanism is to improve the progress of its detection. In the past, it was possible to do a multi-scale fusion pose in the model, but no one integrated the channel attention mechanism into the network.
The third is an intelligent optimization algorithm. For example, particle-based algorithms, there are many kinds of particle and method improvements. For example, one day you suddenly saw someone using a chaotic algorithm in a document, and others quoted chaos into the algorithm. Chaos is used in particles and algorithms. If you take this idea into this aspect and propose a chaotic particle swarm algorithm, it also improves the speed and gets rid of local shrinkage. In this way, you have innovation.
The improvement of the existing model is not behind closed doors. After you find the entry point of your algorithm, you have a systematic and framework understanding of the whole subject, and when you are looking for innovations, you should research some other documents with a purpose. Sometimes you use this idea in this method, and you apply it to yourself. Can this method work on the current topic? I have done this for a few days, has anyone made it? If not, you can boldly say that this is your innovation.
2.2 Application Innovation
The second type is applied innovation. Model creation is difficult, what should I do at this time? The second way of thinking is applied innovation. For example, the first one is an existing algorithm model. The application in psychology is the emotional algorithm mentioned just now. It may have been applied in other fields. You may be in a certain field. One day, in a CNKI or Google Scholar survey, I suddenly saw that someone was applying this algorithm in a certain field. You used it very well, but after thinking about it, the algorithm has never been used in mine. I used it for the first time It is applied to nonlinear control systems or nonlinear control systems, for example, and has achieved a good effect. This is innovation.
3. The journal selection determines the fate of the submission!
Some postgraduate students may wish to publish the Chinese core, and then you can post one or two innovations. If you want to release the EI core, then you must have two innovations. For example, a new combination of several existing algorithm models, if you want to post EI or SCI of area B, then you have to use a new combination of new models that no one else has used for human object detection and pose estimation. There are some minor improvements to the model.
If you want to launch high-level SCI or top conferences, then in terms of principle innovation, you have to prefer how to propose a semi-new or brand-new model. For most of the postgraduate students, it is not the first goal to issue a top conference. Maybe they can graduate with one SCI at most, and the first goal of most people is to graduate smoothly. On the basis of being able to meet the requirements of a smooth graduation, it is possible to go up a little further and post a higher area to provide some foundation for future job hunting, postdoctoral work or research, which will be better. But now all walks of life are very busy. Staying in school for half a year or a year is exhausting and meaningless. It is a serious matter to hurry up to the society and work as soon as possible, because it is getting more and more difficult to find a job every year. So from the beginning to the present, I have always emphasized two words, fast, which means efficiency. The master-doctorate stage is really fleeting. If you don’t pay close attention to the front, you will be very passive later, and then add the pressure of graduation. At that time, your movements will be deformed, so everyone should try to take shortcuts and avoid detours.
3.1 Precautions for selection of submissions
When all these are done, the next step is to face the problem of submitting the manuscript. Many students feel that they can just find a journal to submit their manuscript and submit it, or search the Internet and see that the journal matches my field, so I just submit it casually, but often the selection of journals is very important, because you Choose this journal, maybe this journal rejects you, and you will be selected for that journal. Waiting for you to be rejected by this journal before submitting to that journal, a few months will pass in a while, and the time can't afford to wait, because we don't want to postpone, we all want to graduate quickly, so we must hurry up and select journals at this time It is very important. When choosing a journal, you should pay attention to the following four points and ask yourself,
The first is whether the academic field and direction match? Does the journal you are going to vote for match the direction of the paper you are writing? At this time, you have to look for some historical documents published by them before, to see if there are any articles written in your similar field or direction, which are similar to yours;
The second is to go to their official website to see which fields or directions of papers they accept and match you, then you can submit. Otherwise, you will fail the preliminary review, and it will be a waste of time to return it anyway, half a month. The month passed at once, and time really passed too quickly.
3.2 Ways to select periodicals
The second is whether the level of your academic journals matches the quality of the academic papers you write. This is very important. Before you submit your submission, you should have a pre-judgment on the journals of your own level. How to pre-judge? Or find this journal, for example, your target journal, or there are a number of target journals you have initially selected. They have published papers similar to yours in history, and take them out to make a quality match with papers in your field. How do you think the methodological innovations of your papers, including the style of writing, including the entire logic of writing, compare with others? Is it equivalent? After the comparison, you will have confidence. If other people's method innovation is obviously higher than your level, and now it is more taboo to submit more than one manuscript, and you are sent back to this journal, then your first three or four months may be gone again.
The third is whether the submission cycle of academic journals is appropriate. For example, some university journals often take a year or a year and a half to review manuscripts. For example, when you just entered the postgraduate stage and collaborated with your seniors to publish a paper, at that time you submitted to such a periodic journal. But in the next few years, you will be faced with publishing papers to meet your graduation requirements. You have to think about whether papers with such a long review period are appropriate.
The fourth is whether the 100-point system for academic journals is acceptable. As far as I know, some colleges and universities require masters to publish academic papers, but some people in the project team are relatively poor. Sometimes you have to think about whether you can accept tens of thousands of page fees.
So how do I see if these four points match my current requirements?
I mainly have three ways, the first is to enter the literature citation, the journal to which the literature belongs, that is, when you do your specific method research after getting started, they are in the introduction or literature review stage, they have cited so many literatures, so many The working method will write the title of the journal in a book, you know from that place, it turns out that my field can still publish this, and this journal can also be published. You come to find some journals through this way. You may not know about journals before. Through literature research, you know that there are such journals, and the papers I write can be submitted to this journal.
The third recommendation from the same discipline, for example, your senior sister or classmate has already submitted to this journal, which is very useful. How about its submission cycle? They told you that there is no problem, and now it’s time to go back to inheritance research. They can tell you about this, and they can really help you, whether it’s getting started, finding entry points, methodological research, including In the paper submission, many students are really doing pioneering research on many things, so it is difficult to get started, confused, and confused until the end, they become more and more nervous.
4. How to improve the level of essay writing?
At the stage of my life, my classmates and I regarded publishing papers as a kind of graduation goal, and I regarded it as a kind of work, but in fact, when I really do a postdoctoral fellowship, including I may have to do this in the future For scientific research work in the direction, you will feel that thesis writing is really important. Some students may think that our science and technology group has money, so we can find editors, and the tutor can help solve the remuneration for editors, so should I not improve my thesis writing level? I think dissertation writing is not only for you now, ordinary people may have to do post-docs in the future, and after doing post-docs in China, they have to write notebooks and apply for projects, and some students may want to go abroad for further study, and they are faced with writing RP, or Write an application. At this time, if you can improve your thesis writing level through the limited writing experience of these few papers during your postgraduate stage, and carry out scientific research on you, including applying for topics later and applying for a doctorate abroad, They are all very helpful, so today I want to talk about this point alone.
The first point is that I think translation tools can be used well, such as Google Translate and Baidu Translate. With the improvement of AI translation technology, Google Translate and Baidu Translate’s English-to-Chinese translation has become more and more accurate. If you can use such tools when writing English journals, you can improve translation efficiency.
4.1 Make good use of translation tools
For example, the result of Baidu translation in the picture, there are some bilingual example sentences below according to the keywords. It is sentence by sentence, the logic is not very good, and the expression is relatively simple, but you can use the translation tool to make a translation model for you, and then you can adjust the word order on the basis of the translation, which can save the energy of translation.
4.2 Learn to imitate
The second is to learn to imitate, imitation is the only way to learn. When some students research other people's literature, they may use a translation tool to translate English to Chinese on the basis of others. They only look at how other people's methods are innovative, and then look at what other people's ideas are, and then pass it.
Often like some high-level journals, their writing logic, including their grammatical expressions, are really worthy of our study, such as some as shown in the figure, as shown in the table, as shown in the table, and Below, what does it represent? These collocations can be learned. How do other people’s journals make these collocations? In some journals, passive sentences are mostly used. You may find that high-level journals like to use passive voice. Then I will take the initiative to use more passive sentences when writing papers.
4.3 Make good use of conjunctions
You may find that some high-level SCI writing uses short sentences, and rarely uses long sentences. Long sentences may be too long, which makes it difficult to read. Maybe everyone is used to using attributive clauses, several attributive clauses or clauses One set is very long, but it is often recommended to use short sentences in real SCI writing, and then use logical relative words to link between short sentences. This is the third point, that is, it must be written in English. Make good use of conjunctions to enhance the logic of your writing. The logic of writing just reflects how you express how you solve this academic problem, how you put forward the method, and the logic of your thinking about the problem can actually be reflected in the logic of your writing. So make good use of logical connectives in writing, such as turning, progression, and juxtaposition. When using transition, don't always use however, so many words can be used.
4.4 Synonym Replacement
You may have learned it in high school, college, or postgraduate entrance examination. Don't always use the same similar words, but use more words that can reflect the diversity of your expression. For example, the above turning point, however, never than less, while, these can all be used.
The second is progressive, and progressive is not only and, you can also use further more, what's more, and so on. In academic paper writing, there are quite a lot of progressive relationships.
There is also a causal relationship and a parallel relationship. For example, in the example I gave on the right, in an abstract, there are logical relative words between each sentence, which reflects the entire sentence you want to express. Let the foreigner or Some reviewers enjoyed reading it. Some people's papers may be confusing to read, so you need to pay attention when writing.
The more you get to the end, the more you find that the field will be more and more, and the more you do it, the more you will be free. It is difficult for one person to understand everything very well. If you have a deep understanding, you can only say that you understand roughly, so some students You can follow the process I just mentioned, find the entry point, find your way to get started, and then conduct research.
When you are doing research, try to take the initiative to understand some new concepts. Don’t just do algorithms or do some simulations just to publish articles. To be honest, although you have escaped, it is really difficult for you to find a job. Now I think doing AI is simply doing AI algorithms. You can use the models over there to improve it. It may be no problem to publish articles, but I think the competition will still be quite big in the future.