Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
Dieses Kapitel geht den häufigen Fehlern und korrekten Denkweisen beim Lesen und Schreiben von Einstellungen unter Doktoranden nach. Er hebt fünfzehn Arten von Fehlern hervor, wie etwa alles zu akzeptieren, was in veröffentlichten Zeitungen gelesen wird, ohne hinterfragt zu werden, jede Zeitung Zeile für Zeile zu lesen und falsche Fragen zu stellen. Der Text betont die Bedeutung kritischen Denkens, selektiven Lesens und regelmäßigen Schreibens, um die Forschung voranzutreiben. Es bietet praktische Beispiele, einschließlich der Entwicklung von Techniken zur Visualisierung der Kerndichte, um zu veranschaulichen, wie Forschungslücken identifiziert und ein sinnvoller Beitrag geleistet werden kann. Das Kapitel behandelt auch die üblichen Ausreden für das Nichtschreiben, wie etwa auf Inspiration zu warten oder eine perfekte Idee zu haben, und bietet Strategien, um diese Barrieren zu überwinden. Er schließt mit der Betonung der Wichtigkeit regelmäßigen Schreibens und der Rolle der Vorgesetzten bei der Anleitung der Studenten, ihre akademischen Schreibfähigkeiten zu entwickeln.
KI-Generiert
Diese Zusammenfassung des Fachinhalts wurde mit Hilfe von KI generiert.
Abstract
In this chapter, we discuss some common mistakes for reading and writing attitudes that have been made by new postgraduate students (including us in the early stage of career), which have been categorized into the following fifteen types.
In this chapter, we discuss some common mistakes for reading and writing attitudes that have been made by new postgraduate students (including us in the early stage of career), which have been categorized into the following fifteen types.
4.1 Accept Everything They Read in a Published Paper
Since the undergraduate study in computer science and its related fields is mainly course-based, the contents of those course materials/textbooks are normally correct. The main reason is that those knowledges have been well established/tested in several decades. As an example, Dijkstra’s shortest path algorithm, which is taught in Algorithm courses, was developed in 1957. This algorithm has been used for more than 60 years, which is already a de facto standard. Therefore, the undergraduate students usually regard everything that they read to be the truth. What they need to do is to memorize those materials and to understand the mechanism behind those algorithms/frameworks/concepts in order to get a high grade in each course. They seldom raise some questions (or doubts) about the contents of those course materials/textbooks because these actions cannot help them get the high grades. However, some students (especially for those with high grades) may transfer these mindsets for the postgraduate studies. They may accept everything they read in a research paper, regard it as the truth, and do not raise any question about it. Note that this mindset is completely wrong in postgraduate studies. Here, we use an example for illustrating why it is wrong.
Anzeige
Example: In 2019, the first author of this book worked on a problem called kernel density visualization (KDV), which is used to generate a hotspot map based on some location data points (see Fig. 4.1). He developed efficient algorithms for generating an approximate KDV since the time complexity of this operation is very high, which takes O(XYn) time (where \(X\times Y\) and n denote the resolution size and the number of location data points, respectively). This work has been published in SIGMOD 2020, which is stated below.
\(\bullet \)Tsz Nam Chan, Reynold Cheng, Man Lung Yiu. QUAD: Quadratic-Bound-based Kernel Density Visualization. SIGMOD 2020.
Fig. 4.1
Generating a hotspot map (based on KDV) for the Hong Kong COVID-19 location dataset. (Obtained from Fig. 1 in “Tsz Nam Chan, Leong Hou U, Byron Choi, Jianliang Xu, Reynold Cheng. Large-scale Geospatial Analytics: Problems, Challenges, and Opportunities. SIGMOD Conference Companion 2023”)
Suppose that a MPhil/Ph.D. supervisor assigns this topic for the student to follow up and the student still adopts the mindset for accepting everything he/she reads in a published paper. This student will immediately give up this topic when he/she reads the abstract of this paper (see Fig. 4.2). The most “proper reason” from this student is that this method can already achieve the real-time performance (0.5 s) for generating KDV under a single machine setting (underlined in red in the abstract), which is too good and there will be no room for improvement (i.e., don’t waste time for investigating this topic). If it were really true, the following list of further research studies, which are all in top-tier database and data mining venues (SIGMOD, SIGKDD, VLDB, and ICDE), will no longer appear.
Fig. 4.2
The abstract of the SIGMOD 2020 paper. (Obtained from “Tsz Nam Chan, Reynold Cheng, Man Lung Yiu. QUAD: Quadratic-Bound-based Kernel Density Visualization. SIGMOD 2020”)
Tsz Nam Chan, Zhe Li, Leong Hou U, Jianliang Xu, Reynold Cheng. Fast Augmentation Algorithms for Network Kernel Density Visualization. VLDB 2021.
Tsz Nam Chan, Pak Lon Ip, Leong Hou U, Weng Hou Tong, Shivansh Mittal, Ye Li, Reynold Cheng. KDV-Explorer: A Near Real-Time Kernel Density Visualization System for Spatial Analysis. VLDB 2021. (Demo track)
Tsz Nam Chan, Leong Hou U, Byron Choi, Jianliang Xu. SLAM: Efficient Sweep Line Algorithms for Kernel Density Visualization. SIGMOD 2022.
Tsz Nam Chan, Pak Lon Ip, Leong Hou U, Byron Choi, Jianliang Xu. SAFE: A Share-and-Aggregate Bandwidth Exploration Framework for Kernel Density Visualization. VLDB 2022.
Tsz Nam Chan, Pak Lon Ip, Leong Hou U, Byron Choi, Jianliang Xu. SWS: A Complexity-Optimized Solution for Spatial-Temporal Kernel Density Visualization. VLDB 2022.
Tsz Nam Chan, Pak Lon Ip, Kaiyan Zhao, Leong Hou U, Byron Choi, Jianliang Xu. LIBKDV: A Versatile Kernel Density Visualization Library for Geospatial Analytics. VLDB 2022. (Demo track)
Tsz Nam Chan, Leong Hou U, Byron Choi, Jianliang Xu, Reynold Cheng. Large-scale Geospatial Analytics: Problems, Challenges, and Opportunities. SIGMOD 2023. (Tutorial track)
Tsz Nam Chan, Rui Zang, Pak Lon Ip, Leong Hou U, Jianliang Xu. PyNKDV: An Efficient Network Kernel Density Visualization Library for Geospatial Analytic Systems. SIGMOD 2023. (Demo track)
Tsz Nam Chan, Rui Zang, Bojian Zhu, Leong Hou U, Dingming Wu, Jianliang Xu. LION: Fast and High-Resolution Network Kernel Density Visualization. VLDB 2024.
Tsz Nam Chan, Pak Lon Ip, Bojian Zhu, Leong Hou U, Dingming Wu, Jianliang Xu, Christian S. Jensen. Large-scale Spatiotemporal Kernel Density Visualization. ICDE 2025.
Yue Zhong, Tsz Nam Chan, Leong Hou U, Dingming Wu, Wei Tu, Ruisheng Wang, Joshua Zhexue Huang. A Fast and Accurate Block Compression Solution for Spatiotemporal Kernel Density Visualization. SIGKDD 2025.
Therefore, we can observe that this student may give up the chance for establishing a new field and give up a lot of top-tier publications if he/she adopts the wrong mindset by accepting everything he/she reads in a published paper. Here, we would like to emphasize that many authors would like to use the style of “selling the good things and hiding the bad things” for writing papers. The main reason is that this approach easily makes reviewers feel impressed so that they can give an “Accept” for their papers. As such, the readers need to ask questions/raise doubts about what they read instead of accepting everything. Asking questions (see Fig.4.3) is the only way to (1) figure out the weakness of the paper and (2) identify the new research directions.
Fig. 4.3
Unproductive students do not ask questions when they read papers, while productive students ask questions and mark down many new directions when they read papers
When the students are in the undergraduate studies, they need to read the course materials (or probably textbooks) from cover to cover in order to grasp all those concepts so that they can get an A+/A grade in each course. These students may keep the habit and read every paper line-by-line and cover to cover. They may think that this can help them understand the paper thoroughly. However, this mindset is not correct, which is not realistic and feasible for obtaining/understanding knowledge nowadays. In recent years, gradually more papers are submitted to top-tier conferences/journals. Consider several top-tier AI/ML conferences in 2024, including IJCAI, AAAI, ICML, and NeurIPS. Each conference accepts more than 1000 papers in that year. Using Fig. 4.4 as an example, the number of pages of the last paper in AAAI 2024 has already reached 23,861. Some database conferences, which have relatively smaller numbers of papers for each year, also contain more than 200 publications per year. As an example, the number of research papers that are presented in VLDB 2024 is more than 250.
We expect that the numbers of publications of all conferences/journals will continue to increase in the future since (1) more undergraduate students would like to undertake the postgraduate studies and (2) some subfields in computer science, e.g., artificial intelligence (AI), computer vision (CV), natural language processing (NLP), deep learning (DL), large language model (LLM), and data science (DS), start affecting other subjects (e.g., chemistry and physics (AI for science)), which can further arouse the interest from more students to join.
Furthermore, reading a paper line-by-line (and cover to cover) does not mean the reader can understand that paper. Some papers (especially for theoretical papers) involve a lot of details (e.g., cumbersome proofs), which can possibly distract the reader. Instead, the reader may understand more if they can summarize the paper in a high-level way1 (but not include many details).
Therefore, instead of thoroughly reading a lot of (probably not important) papers, a student needs to understand which parts of each paper should be important for reading and which parts should be skipped. Here, we illustrate how to adopt the inverted triangle approach to filter those papers (see Fig. 4.5). Note that the student can already filter a lot of papers based on reading the titles. As an example, suppose that the student wants to conduct research related to spatiotemporal data management. He/she can skip those papers related to graph data management or relational data management. Once the student identifies some attractive papers based on the titles, he/she can directly read abstract/introduction of these papers during a short period of time (maybe a coffee break). With this short period of time, he/she can further identify some important papers. As an example, suppose that the student develops interest in working on kernel density visualization. He/she can skip those papers related to other types of visualization (e.g., scatter plot) after he/she reads the abstract/introduction. After the student has found that the title and abstract/introduction are interesting (and are worth for reading further), he/she can read the preliminaries and related work by identifying some missing research studies (e.g., there are other missing papers that are related to kernel density visualization). Only a few important papers (e.g., direct competitors) need to be thoroughly read by the student. Here, we further consider the first author of this book as an example. When he worked on the research paper “KARL: Fast Kernel Aggregation Queries”, which has been published in ICDE 2019, he only read three papers thoroughly (and read many papers partially), which are the following three papers.
Edward Gan and Peter Bailis. Scalable Kernel Density Classification via Threshold-Based Pruning. SIGMOD 2017.
Alexander G. Gray and Andrew W. Moore. Nonparametric Density Estimation: Toward Computational Tractability. SDM 2003.
Thomas Seidl and Hans-Peter Kriegel. Optimal multi-step k-nearest neighbor search. SIGMOD 1998.
Based on the above discussion, we can know why it is impossible and not wise for researchers to read every paper line-by-line (and cover to cover). Remember that a very good and productive researcher only fully reads a few papers.2
Fig. 4.5
The number of reading papers should follow this inverted triangle
Many students (especially for junior students) may say that they are hard to find their own research problems although they have read many research papers and have asked a lot of questions regarding those papers. Some students may even mention that they struggle to find a research problem for nearly a year. However, after we investigate how they read research papers, we discover that most of these students may directly dig into the details without understanding the background of their research problems. Using Fig. 4.6 as an example, this student may want to find one research topic to work with. As such, he/she searches for the dblp record and (randomly) finds out one SIGMOD paper. However, he/she may skim-read or even skip the abstract and introduction and directly goes through the technical sections. It is very natural for many computer science students to do this because they may think that they are technical guys, who are good at developing algorithms but do not need to listen to a story. Therefore, they may ask a lot of questions about the details (e.g., “How to derive Eq. 14?” in Fig. 4.6). However, these questions are unlikely to help them find new research problems/directions. Instead, they should carefully read the abstract and introduction so that they can easily ask some questions that can open up a new direction. Using Fig. 4.1 as an example, when the authors have mentioned that they can improve the efficiency of generating a KDV-based hotspot map in abstract and introduction, the readers can ask some questions regarding the problem settings.
(1) Can we improve the efficiency of supporting the spatiotemporal hotspot map?
(2) Can we improve the efficiency of exploratory operations (e.g., zooming and panning) for the KDV tool?
Fig. 4.6
Many junior students can ask incorrect questions for finding new research topics when they are reading papers
As a practical example, we illustrate how the first author of this book found new research topics from September 2017 to April 2021 by asking correct questions in Fig. 4.7. Note that those references from [a] to [i] in the figure are shown as follows.
a
Tsz Nam Chan, Man Lung Yiu, Leong Hou U. KARL: Fast Kernel Aggregation Queries. ICDE 2019.
b
Tsz Nam Chan, Reynold Cheng, Man Lung Yiu. QUAD: Quadratic-Bound-Based Kernel Density Visualization. SIGMOD 2020.
c
Tsz Nam Chan, Leong Hou U, Reynold Cheng, Man Lung Yiu, Shivansh Mittal. Efficient Algorithms for Kernel Aggregation Queries. IEEE TKDE 2022.
d
Tsz Nam Chan, Zhe Li, Leong Hou U, Reynold Cheng. PLAME: Piecewise-Linear Approximate Measure for Additive Kernel SVM. IEEE TKDE 2023.
e
Tsz Nam Chan, Pak Lon Ip, Leong Hou U, Weng Hou Tong, Shivansh Mittal, Ye Li, Reynold Cheng. KDV-Explorer: A Near Real-Time Kernel Density Visualization System for Spatial Analysis. VLDB 2021 (Demo track).
f
Tsz Nam Chan, Zhe Li, Leong Hou U, Jianliang Xu, Reynold Cheng. Fast Augmentation Algorithms for Network Kernel Density Visualization. VLDB 2021.
g
Tsz Nam Chan, Pak Lon Ip, Leong Hou U, Byron Choi, Jianliang Xu. SWS: A Complexity-Optimized Solution for Spatial-Temporal Kernel Density Visualization. VLDB 2022.
h
Tsz Nam Chan, Pak Lon Ip, Leong Hou U, Byron Choi, Jianliang Xu. SAFE: A Share-and-Aggregate Bandwidth Exploration Framework for Kernel Density Visualization. VLDB 2022.
i
Tsz Nam Chan, Leong Hou U, Byron Choi, Jianliang Xu. SLAM: Efficient Sweep Line Algorithms for Kernel Density Visualization. SIGMOD 2022.
Based on the above discussion, we further emphasize that only the above question type (related to problem settings) can significantly help students find new research directions. Therefore, those students should frequently ask this question type (but not to ask for the details) when they are searching research problems.
Fig. 4.7
This is how the first author of this book asks correct questions to discover new research directions from September 2017 to April 2021
Many students may think that conducting computer science research is to (1) learn state-of-the-art technology from research papers, (2) develop new algorithms/systems, and (3) writing code. They may think that writing a paper should be an easy task, which can be done in a few days, if all those new solutions/techniques are available. Therefore, they may imagine that all they need to do is to sit down, drink a coffee, and think of new solutions/techniques. When the supervisors ask them to write somethings, they may think that this is an interruption of their thoughts (just like interrupting Albert Einstein). Therefore, they may arbitrarily write somethings and return them to their supervisors. Ultimately, these students will find that they have zero (or nearly no) progress for their research papers and observe that their colleagues (with correct mindsets) have already submitted 5 to 6 papers to (and with one to two papers accepted in) top-tier venues in a year. At that time, some of these (especially for those cocky) students may think that the productive colleagues may just write some “water” papers (This word comes from Chinese, meaning that those papers are extremely incremental and not important), which cannot advance the state of the art. They will possibly keep this habit and think that they will ultimately have some big ideas that can change the world. But before these ideas come true, they may finally graduate with no paper (or with some weak papers) so that no university wants to hire them (i.e., leaving the academia forever). Even worse, some of them may be kicked off from research labs and never graduate with MPhil/Ph.D. degrees.
In addition, some students may think that their MPhil/Ph.D. supervisors will help them edit those research papers. Therefore, they will not treat those writing tasks (assigned by their supervisors) in a very serious way. They may think in this way. “My supervisor will help me revise the paper. Let me arbitrarily write it and send it to my supervisor.” If my students really do this kind of things to me, I will blame them seriously. First, the supervisor needs to handle a lot of issues (e.g., writing proposals, writing books, writing research papers, writing patents, teaching courses, preparing teaching materials, attending conferences, attending MPhil/Ph.D. defenses, and having administrative meetings), which can possibly be ten times busier than those students. Second, the supervisor should have no responsibility for the MPhil/Ph.D. degree given that the student is lazy. Third, students can no longer learn the writing skills and earn the writing experience if they do not care much about those writing tasks.
Although learning new technology, developing algorithms/systems, and writing code should be the important components of conducting computer science research, writing should be regarded as the most important component of conducting computer science research. If students do not write, no one can know what they think (even they may have big ideas). Therefore, they must treat this to be the serious task during the MPhil/Ph.D. studies.
4.5 Avoid Writing Before the “Inspiration” Comes
Many students may say that they have no inspiration (a.k.a. no idea) for writing. They may also argue that they do not know what to write next. Because of the above issues, the last time for opening latex or overleaf can be several months ago (i.e., they do not edit papers for several months). One example can be found in Fig. 4.8 (in the overleaf). This is how we collaborate with another group of researchers. Note that the last time for updating this paper is four months ago and the first author only has this research task.
Fig. 4.8
There is no update for the paper in four months. Due to privacy issues, we hide some parts with white rectangular boxes
At first glance, it seems that their argument is correct. Suppose that students do not have any idea. What should they write? However, once you ask them about the progress regularly, they will always say somethings similar to you (e.g., They have no idea. The idea is not good. They are now thinking about the idea.). Then, we wonder why they do not have any idea for writing in a long period of time. Finally, we figure out some possible situations for those students.
Some students just sit down in a library/an office and start thinking about ideas. They think that those ideas will suddenly jump out if they sit for a long time. Suppose that you belong to this type of students. Then, we will ask you. Have you ever successfully established the complete solution in your brain? Here is our answer. We reckon that most of them find that their brains are still blank and nothing comes out even though they sit for a long time. Although some of these students can catch some small ideas during a few hours (very lucky), they may forget them after they have done other things (e.g., replying to emails, checking messages from WeChat/WhatsApp, browsing the Internet, chatting with their friends, and buying a cup of coffee). Several months later, they may discover the same idea again (i.e., they do not have any progress in these months and their brains get lost.).
After we discuss the above situations with those students, they totally agree with us and wonder why we know their situations so well. It is because a brain is normally in chaos and it is very easy to be distracted by other things. Using Fig. 4.9 as an example, everyone has a lot of stuffs in the brain. As an example, he/she may be worried about the relationship with the girlfriend/boyfriend. As another example, he/she may also get distracted by the melody of a new song. With a lot of stuffs in the brain, it is very hard to focus on thinking the research ideas. Even though you can discover the new ideas (and is satisfied with the progress for that day), it is easy to lose those ideas (with the reason that the brain is in chaos, which is hard to store complicated stuffs related to research ideas.).
Fig. 4.9
The brain can be in chaos even though you want to think of the new research ideas
Based on the above discussion, we can immediately know that it is very hard to establish a full solution to a research problem in a brain. Moreover, it is also very easy for a brain to forget the idea. Therefore, we cannot simply think, wait for the “inspiration” to come, and then write papers. Instead, we need to write in the very early stage. Note that writing is the creation process, which forces the writer to concentrate on what he/she writes. Therefore, it can help avoiding other stuffs (e.g., girlfriend and songs) to interrupt the writer, which can increase the chance for having inspiration (i.e., new idea). In addition, once the new idea is discovered, the writing process can record this idea so that it will not be (re)discovered again and again. As such, we can say that it is not “inspiration leads to writing”. Instead, it should be “writing leads to inspiration”.
4.6 Avoid Writing Before the Idea is “Perfectly” Tested
Many students may say somethings like this.
(1)
“I do not write the paper because I still need to conduct more experiments for testing my idea.”
(2)
“I need to compare my method with the additional baseline method. Therefore, I do not want to write the paper until I have seen the result.” (Note that he/she has already compared the method with some existing methods and obtained the results.)
(3)
“I need to thoroughly test my method in order to verify whether it can really advance the state of the art in order to determine whether it is worth for me to write this paper or not.”
At the first glance, we may think that these students are very rigorous and are serious about the experiments. However, once we ask them about their progress in the next week (or month/or even year), they will reply you in the same way. Then, we immediately know that they may have either the following three issues.
(1) Excuses for not conducting research. Some students say this because they may only have a rough idea (or even no idea) for how to push the progress of their research. They may have no motivation for searching possible solutions. But they still need to pretend to be hard-working so that other people can admire them. For those students, we would like to say that this is the most childish thing. In reality (of research), only awards, publications, patents, monographs, research grants, and systems, etc., can be counted as research outcomes. No one cares about how they work hard toward conducting experiments, writing code, or testing ideas. Suppose that they would like to waste time for pretending to be “hard-working” students. They should consider whether they can (1) find a way to advance their research or (2) find somethings that are worth for them to do (maybe they can leave academia. Mark Elliot Zuckerberg (the co-founder of Facebook/Meta) also dropped out Harvard University.).
(2) Excuses for not writing papers. Indeed, some students are willing to conduct research. However, they do not have enough knowledge for (or do not like) writing papers. Therefore, they choose to write code/conduct experiments, which should be the relatively easy tasks, in the daily time. For those students, we would like to point out that the less papers they write, the less writing experience they can gain. We know that it is very painful to start writing an article, especially for the first one (the junior students). All of us have this kind of painful experience before since we did not have any knowledge of academic writing when we started the postgraduate studies. However, when you look back to your life, you will realize that you must have other types of painful experience for learning. As an example, when we learned the topic of Mathematical Induction in high school, it also took us a long time to be skillful for using this to prove whether the mathematical statement is correct (by doing a lot of exercises). You also got through it at that time. Therefore, why do you avoid writing academic papers? Note that the academic writing skills, which may be used for your next 30 to 40 years in your academic life, can be more important than Mathematical Induction.
(3) Easy to give up ideas. Some students say this because they really want to thoroughly (or perfectly) test the idea before writing. Note that every idea should have its weakness (i.e., no perfect idea). Otherwise, we can see that each textbook in the computer science field only discusses one (so called the best) solution. In addition, it is also not necessary to have so many researchers in many fields. Therefore, as shown in Fig. 4.10, if students do not write and figure out the weakness of the idea, they may be frustrated and think that they should give up this idea (or even move to the next topic). Why do they easily give up this idea? It is mainly because they have no draft. As such, the students (1) think that it has no cost to move to the next topic, (2) saving the idea is very tired, (3) focusing on the same topic is also very tired for them, and (4) changing to the new (fresh) topic is good. However, if students write regularly, they will have the draft with many pages (e.g., eight pages in Fig. 4.10). Once they figure out the bad performance for their idea, they have the high cost for giving up. What is the reason? Since they have already written many pages (e.g., eight pages), they only need to write a few pages, e.g., four pages (the number of pages for a lot of venues is 12.), in order to submit this paper to a venue. As such, it is not wise (or takes high cost) for them to give up this topic.
Based on the above discussion, we further emphasize that there is no perfect idea. In addition, aiming “perfect” might be the excuse for you not to conduct research/write papers. Keep writing every day.
4.7 Never Write Regularly
Many TV shows describe that the top scientists have this kind of behaviors. First, they sit on a chair and wait for the inspiration (the big idea) comes. Second, they are always “brainstorming” (do everything in their brain). Third, they develop a big physical system that can change the world. Therefore, many new postgraduate students may be misled and think that conducting research is to (1) make themselves clever, (2) wait for their brains to provide novel ideas, and (3) only report when big experimental outcomes are available. However, none of these TV shows mentions that those scientists need to write papers regularly. Worse still, many recent news (especially in China) report that some students may have written a lot of incremental (“water”) papers. Although we also do not encourage students to write these incremental research papers (which cannot benefit their careers), these news may easily be misinterpreted by those students (especially the junior ones) that writing papers are relatively not important.
In fact, writing regularly is a very important step for conducting research. For us, we believe that this should be the most important step (even more important than performing experiments). The main reason is that a draft is a concrete thing, which can show the current understanding of the writers. When you write, this draft can automatically tell you what you should do next (Of course, you need to keep asking questions.). As an example, when you get stuck for writing because you are not very familiar with some parts of related research studies, the draft will tell you to check those related studies. As another example, when you get stuck for writing because some experimental results are not available, the draft will tell you to conduct those experiments.
Many students may still wonder why they need to write papers regularly. They may argue that the brain should be responsible for thinking ideas. However, as we mentioned before (in Fig. 4.9), the brain is normally in chaos, which may contain a lot of stuffs. As such, even though the students may come up with great ideas, they can forget some critical details or even the whole ideas after they have performed other tasks, e.g., checking the WeChat message, gathering, and taking a rest (see Fig. 4.11). In contrast, the students can immediately further elaborate the ideas that they thought before once they have marked down those ideas in the draft (see Fig. 4.12).
Fig. 4.11
Researchers who do not write are hard to have any progress in research
Here, we further use this analogy for illustrating the importance of writing papers. We think that everyone plays role-playing (computer) games before. For each of these games, the main character needs to (1) go to different places, (2) fight with different monsters, and (3) fight with several bosses in order to win this game. Normally, it takes a long time for each player to finish the game. Therefore, each game allows the player to save a record so that the player can start from this record next time. I think everyone must save the record when he/she plays this game. Otherwise, they will start this game from scratch again. In fact, we can regard that writing a paper is the same as saving a record in a game. If we do not write, we need to start thinking ideas from scratch (see Fig. 4.11) as the brain may not be able to store the ideas (given that the brain has a lot of stuffs). If we write, we can start thinking on top of those ideas that have been recorded last time (see Fig. 4.12).
Based on the above discussion, we know that writing regularly is the way to record what we have done in our research, which can significantly help the progress of our research.
4.8 Avoid Writing Because of No Mood
Many students do not write research papers because they have no mood for writing. Indeed, this reason is very natural because academic writing is difficult. The brains from these students can easily turn to blank when they see the blank page of the PDF that is generated by latex. Note that we also had this experience before (even for now), especially for some papers that we do not know how to start writing them. At that time, some unproductive students may say that they have no mood for writing anything and then start doing other things, e.g., playing the online game (see Fig. 4.13). For those students, we would like to ask this question. When will you have good mood for writing? The answer is definitely never. The main reason for them to have no mood is the difficulty of writing papers. However, this type of difficulty exists forever, which indicates that these students will have no mood if they do not go through this type of difficulty.
Fig. 4.13
No mood should not be used as an excuse for not writing
One approach to go through this type of difficulty is to force themselves for writing one to two sentences every day (see Fig. 4.13). Initially, it can be very painful (hard to write even for one sentence). However, if the brain is getting used to it, that student does not feel very painful about this task and can start writing more (e.g., one to two paragraphs). Ultimately, that student may not say that he/she has no mood because he/she automatically opens the latex and writes somethings every day.
Using the first author of this book as an example. When he was in February 2016 to February 2017, he had worked on the following computer science topics (by reading several research papers).
Template matching with deformable templates in pattern recognition (roughly February 2016 to March 2016)
Route planning problems in road networks (roughly early March 2016)
Locality-sensitive hashing (LSH) for similarity search (roughly March 2016 to April 2016)
Document similarity search (roughly April 2016)
Similarity search for non-metrics (roughly May 2016 to July 2016)
Clustering (roughly July 2016 to August 2016)
Manifold learning (roughly September 2016 to October 2016)
Time series similarity search (roughly November 2016 to early January 2017).
However, he always felt no mood (or was not motivated) for writing any research paper. As such, he had not written even one page of a research paper at that period and changed research topics frequently. Once he felt that he would have the graduation issue and had read the slides3 from Dimitris Papadias for stating that “a draft is something concrete–otherwise you may have done nothing as far as I am concerned.”, he started to force himself to write several sentences when he started the new topic of Earth Mover’s Distance similarity search in February 2017 (He sent the first incomplete draft (with 1.5 pages) to his supervisor on 8th February 2017). Initially, it was a painful process for him as he did not write a lot before 2017. However, after several months later, he started to see that the written draft automatically guided him what to write/do next. Therefore, he started to have a mood for opening latex to write several sentences every day. Ultimately, this paper was finished in September 2017 and submitted to VLDB 2018. Although this paper was rejected by this conference, this is the end for him to (1) have no mood for writing, (2) do not know what to do next, (3) have no progress in research, (4) change research topics frequently, and (5) have no research outcome. This paper was later also accepted in IEEE TKDE, which is a prestigious journal in data engineering.
Tsz Nam Chan, Man Lung Yiu, Leong Hou U. The Power of Bounds: Answering Approximate Earth Mover’s Distance with Parametric Bounds. IEEE TKDE 2021.
After he finished working on this paper, he then used nine months (from September 2017 to June 2018) for writing another paper, which was later accepted in ICDE 2019. After graduation, He started writing research papers in a rapid way and published a lot in SIGMOD, VLDB, ICDE, and TKDE (from 2019 to now) as a first author/corresponding author (especially when he was a research assistant professor in Hong Kong Baptist University and a distinguished professor in Shenzhen University). Now, he never says that he has no mood for writing because it is a habit for him to write every day.
4.9 Avoid Writing Because Their English is Not Good Enough for Writing
We believe that this is indeed a barrier for students (especially for those non-native speakers) not writing papers. Some students may not use English frequently and may have a very painful experience for learning this language in primary/high schools. In some countries, including China, Germany, France, South Korea, and Japan, most of those undergraduate students take computer science courses in their own native languages, who cannot have the chance to use English in class. Therefore, they fear for using English to communicate with others, let alone to write a 12-page English article (with double column) in a very rigorous way. Here, we would like to point out our thoughts regarding this issue.
You should be proud of using English. Some students may think that they will mainly reside in their own countries in the future. Therefore, they will raise this kind of questions. Why do they need to learn English? This kind of mindsets is in fact wrong. For us, mastering an important foreign language should be regarded as a glory, which can have a lot of benefits. Using China as an example, everyone can speak Chinese. If you also use Chinese there, no one will think that it is very surprising (see Fig. 4.14). But imagine that you can use fluent English to communicate with others (see Fig. 4.15). We would say that other people will definitely admire you. What is the reason? It is because they do not have this important skill but you have. Once you are proud of using English, you will definitely have the motivation to learn it. As such, this is the most important step toward learning English.
Fig. 4.14
It is not very special for someone (e.g., a Chinese person) to speak his/her native language (e.g., Chinese)
The best way to get rid of fearfulness for writing English is to write it. After the students are proud of using English, the next step is to get rid of the fearfulness of using this language. To achieve this goal, they need to force themselves to write a few paragraphs in English every day. For example, many students need to read papers every day. But most of them simply read those papers but do not write for what they have read. Summarizing papers in their own words (without any copy-and-paste from the original papers) is a great way for improving their English presentation skills. Moreover, many students may also conduct preliminary experiments. However, they do not write anything after they have conducted them. Note that summarizing experiments can also improve their English presentation skills. With more experiences, they can write the technical and introduction parts, which can further help them for gaining English presentation skills. Note that they need to read previous papers from top-tier venues in order to mimic the writing styles during the writing (learning) process. Although this process is very painful in the early stage, they will realize that this situation is getting better and better after a few months/a year (if they insist writing English every day). The main reason is that they get used to it.
Academic English is not that difficult compared with what you think. Some students may think that academic English must be very difficult. Indeed, when they are the newbies, it is very common for them to have this feeling. As an example, some specific terms/vocabularies in the papers may not be easily understood. As another example, their brains become blank (do not know how to start) when they try to write somethings in latex. However, after the students have read more papers, they will find that many similar sentence structures and vocabularies are reused again, which will automatically be stored in their brains. When the students practice more by writing some paragraphs regularly, they will realize that they can write these structures and vocabularies soon. In fact, writing a research paper in computer science does not need to have extremely solid English skills. There is no need for you to understand those books from Shakespeare or understand the law statements. To our understanding, we believe that the students with undergraduate degrees from computer science in many non-native English speaking countries (including China, Germany, France, South Korea, and Japan) also have enough English skills for reading/writing academic papers if they practice more.
The most difficult part for writing is not the language itself. Some students may argue that they cannot explain the concept clearly because they need to use English (the non-native language). However, once we ask them to explain the concept in Chinese (the native language), they still cannot be able to explain it clearly. The main reason is that they cannot understand the concept clearly or they have not learned the correct methodology for presenting the concept so that it is hard for audience to understand. Under this circumstance, what they need to do is not to blame for their “poor” English. Instead, they need to further understand this concept or learn for the presentation/organization skills. Nowadays, there are many software tools, including ChatGPT and DeepSeek, that can help for improving the language of writing. Therefore, other parts, e.g., the presentation flow, can be more important compared with the language itself.
Based on the above discussion, we need to encourage some students to be proud of using English. Moreover, we also want to highlight that academic English is not very difficult. By reading and writing more, they can ultimately get rid of fearfulness for using this language to publish research papers. Note that practice makes perfect.
4.10 Avoid Writing Because the Initial Draft Will be Ultimately Erased
Many students (especially for junior postgraduate students) may have this kind of experience before. He/she writes the draft and sends it to his/her supervisor. Then, the supervisor can have a lot of suggestions for the draft so that it needs to be significantly revised. More specifically, the student may have spent a week for writing 100 sentences and drawing two figures in an initial draft. In the next version, as suggested by the supervisor, 80 sentences and these two figures are significantly revised. Some of them may even be removed. Based on this experience, many students may think that this process is very painful since they have done a lot of “useless” work (i.e., “waste” a lot time). Some students may even think that they should not write even a single word until they have the “perfect” plan for writing (in order to avoid “useless” work). For those students who do not write any word, we will ask them one question. Have they ever thought of the perfect plan so that they can write the paper in one shot? If yes, how much time do they spend for writing one paper? For my opinion, I think most of them may have never successfully written a complete paper. Even though they have (luckily) completed one paper, they must have spent a lot of time (much more than the time for those students who write regularly).
In reality, it is very hard for everyone to know the next step if we do not try. Our brain cannot predict too many steps especially for the lack of information. Using the maze (see Fig. 4.16) as an analogy of writing papers, it is nearly impossible for the player to predict the correct route in the starting point of the maze. Therefore, the only way to win this game is to adopt the trial-and-error approach to walk through different routes. Although it can be painful for the player to choose the wrong routes (e.g., routes 1, 2, and 3 in Fig. 4.16 can reach the dead end.), these (wrong) routes can provide additional information (i.e., hints) for guiding you to choose the correct route (e.g., route 4 in Fig. 4.16). Therefore, only the players who try can be successful and earn additional experience (and can be confident for playing the next (probably more complicated) maze).
Note that writing a paper is similar to walking through a maze. Since the student may not be very familiar with the topic initially (i.e., does not have enough information) and does not have enough writing experience, it is very easy for him/her to make mistakes for writing a paper (i.e., walk to the dead end). However, during the interactive process with his/her supervisor, he/she can learn more (i.e., gain more information and experience) so that his/her draft will be improved. As an example in Fig. 4.17, the 1st version can be very bad, which can be full of mistakes and the solution is not well-designed. But we need to note that the following versions can be continuously improved. What we mean is that writing the “very bad” version is not a waste of time. Without this version, it cannot guide you to have the next (better) version, and thus is impossible to achieve the perfect version (i.e., the 6th version in Fig. 4.17). Note that writing a paper is also similar to eating bread. We are full because we have eaten six loaves of bread. But it is not because we eat the 6th loaf of bread.
Fig. 4.17
Productive students write a lot of versions of their papers, while unproductive students simply “think” to achieve the perfect version
Given the above discussion, it is nearly impossible for a student to write a perfect paper in one shot. A productive student writes a lot of versions of his/her paper, where each version is improved compared with the previous version, until that version is good enough for submitting to a top-tier venue.
4.11 Avoid Writing Because of Busyness
Many students may say that they do not write because they are very busy with other things, e.g., coding, thinking ideas, working on assignments, having a tutorial class for students, and meeting with others. For those students, we would like to emphasize that this reason can possibly be an excuse for not writing papers (see Fig. 4.18).
Fig. 4.18
The word “busy” can possibly be an excuse for not writing papers
The major goal for postgraduate students is to produce research outcomes. Many students may have a lot of activities every day. Some of these activities may not be related to research. Those students have many activities because they have a lot of free time during the postgraduate studies (Note that it is not mandatory for students to stay in offices during the working hours.). Therefore, it is easy for those students to have bad time management (or become lazy for research). For example, it is possible for students to be busy with playing online games and watching movies during the working hours (can even be from day to night). Although students can enjoy the free time (we regard it as a kind of academic freedom) during the postgraduate studies, they need to understand that they have the responsibility for producing research outcomes (research papers). Therefore, students are expected to assign a large portion of time for conducting research/writing research papers every day. They should not use the word “busy” as the excuse for doing somethings that are not research-related.
Students do not need to have a big block of time for writing. Some students may sometimes need to deal with a lot of trivial matters. Common examples include conducting tutorials for class, marking assignments, attending classes, meeting with others (e.g., supervisors and collaborators), and dealing with some matters from their families. Therefore, they mention that they only have the fragmented time so that they cannot work on anything. Some of them may even emphasize that the fragmented time is so little that they cannot turn themselves into the working mode. As such, they would argue that they can be productive when they have a big block of time (e.g., summer and winter holidays) for writing. However, when we dig into the details, we discover that those students still have at least one to two hours for the fragmented time every day, which are much more than enough for writing. Consider the first author of this book as an example. When he is a faculty member in Hong Kong Baptist University or Shenzhen University, he needs to (1) conduct lecturers, (2) prepare slides for classes, (3) attend meetings, (4) attend academic conferences, (5) give talks in different universities, (6) review papers from academic conferences and journals, (7) become the organizing committee members in some conferences, (8) make questions/answers for assignments/examination papers, (9) deal with teaching issues from students, (10) meet with his postgraduate students, (11) attending defenses from undergraduate/master/Ph.D. students, (12) handle the reimbursement and purchasing issues, and (13) deal with administrative issues from the university. Therefore, his daily time can be much fragmented compared with postgraduate students. However, he utilizes those fragmented time in a much better way. Note that he continues to push the paper/book/patent application even though the fragmented time is smaller than an hour (or even half an hour). As an example, he took a flight from Hangzhou back to Shenzhen on 30th November 2024. At that time, he waited for the plane (with 30-min fragmented time), he still managed to finish the patent application form there. As another example, he taught two courses, attended several defenses, and dealt with other issues in the spring semester of 2024/2025, he still managed to write this book during his fragmented time. Therefore, we reckon that there is no need to have a big block of time for writing. In fact, some other books, e.g., “How to Write a Lot: A Practical Guide to Productive Academic Writing” by Paul Silvia and “Writing Your Journal Article in Twelve Weeks” by Wendy Laura Belcher, also emphasize that it is not necessary for writers to have a big block of time to write in order to be productive.
“Coding” may be useless without writing. Some students may say that they cannot write because they are busy with writing code. For those students, we would like to ask this question. Do you find that coding can help you push the progress of your research work? If your answer is no, we would like to say that you simply write code without any goal. Consider the first author of this book as an example. He worked on one research topic related to improving the efficiency of solving the kNN search with non-metrics from May 2016 to July 2016. At that time, he wrote a lot of code for testing different methods. However, he only regarded writing code as the goal and did not know what to do next after he finished this task. For example, he did not provide any comprehensive analysis for results, did not ask further questions, and did not write research papers. Therefore, his code and his results are useless since they do not further push his research. Here, we emphasize that coding without writing can be useless (see Fig. 4.19). Instead, the best way for having progress in research is to let writing drive for other tasks (see Fig. 4.20), including coding.
Fig. 4.19
Coding without writing can be useless for pushing the research progress
4.12 Never Write Seriously Because my Supervisor Will Help me
Many students think that their supervisors will help them revise papers. Therefore, they are not very serious about their writing tasks, which are assigned by their supervisors (see Fig. 4.21).
Fig. 4.21
You can only gain the writing experience when you write seriously
Here, we emphasize that this mindset is completely wrong. Consider the first author of this book as an example. When he was a Ph.D. student, he conducted research related to template matching from September 2014 to February 2016. At that time, he also had this kind of mindset (i.e., did not seriously write papers). Therefore, although his supervisor had helped him rewrite two papers, which were published in SSTD 2015 and TKDE 2017 (a top journal), he did not gain a lot of experience about how to write a research paper. Worse still, he did not learn how to ask research questions (and thus the problem finding skills). Therefore, he suffered a lot from February 2016 to February 2017 (see the example in Sect. 4.8). As another example, we have also heard from some students who published many decent papers during the MPhil studies cannot publish research papers when they are in the Ph.D. studies from another university. The main reason is that their MPhil supervisors are very nice for helping them rewrite the paper so that they are unable to learn the skills.
Based on the above discussion, we would like to point out that only those students who seriously write research papers (go through a lot of painful issues by themselves) can gain experience so that they can learn how to independently conduct research. Otherwise, they can only be under the umbrella of their supervisors. Once they leave their research groups, their productivity can sharply drop because they cannot get any help from their supervisors anymore. Note that another situation is that some supervisors are not very nice (or very busy). If their students cannot write research papers by themselves, they will be kicked out from the research groups or cannot graduate from the universities on time until they finish enough paper submissions/publications.
4.13 Blame Others for Not Understanding Their Drafts/Presentations
When students send their drafts to their supervisors, those responsible supervisors will read them and provide comments/additional questions to the students. However, some students may not be very happy about the thoroughly edited drafts (e.g., see a big cross for a paragraph and see many questions in the blank space) because they think that they have spent a lot of time for writing the drafts. Some of them may think that the supervisors are so stupid (see Fig. 4.22), who cannot understand (1) the “beauty” hidden in the draft and (2) how genius they are. These students may even question the ability of their supervisors (e.g., think that the supervisor cannot help them become the next “Albert Einstein”).
Fig. 4.22
Unproductive students normally blame supervisors for the poor presentation of their papers
For the above cocky students, they normally have no publication or very weak publications (e.g., papers in third-tier venues) because of the following reasons. First, these students still do not realize that this is the fault of the authors (not readers) if the readers cannot understand their papers. Second, the supervisors guide them for writing well-structured papers. However, the students refuse to learn it, indicating that these students cannot improve their presentation ability to the level of top-tier venues. Therefore, when they want to judge the ability of their supervisors, we would like to say this to them. “Go to the personal homepage of your supervisors and check how many top-tier publications they have before you judge their ability. Then, go to your personal homepage and see how weak you are.”
Compared with the above unproductive (and cocky) students, some productive students are happy to accept comments from their supervisors so that they can learn from the comments for revising their papers. They will even take a look for those old papers from their supervisors in order to learn the writing style. With the training for a few years, they will discover that their abilities (including presentation skills and topic-finding skills) become closer to the ones of their supervisors. With more practices (by writing more papers), they will find that their supervisors spend less time for assisting them, indicating that they are stronger. Ultimately, these productive students will find that they can surpass their supervisors (see Fig. 4.23), who can conduct independent research.
Fig. 4.23
Productive students normally follow the suggestions (especially for those presentation-related suggestions) from their supervisors and improve their drafts so that they can ultimately surpass their supervisors
Many students (especially for junior students) do not understand research papers after they read them. As such, they may then develop this mindset (see Fig. 4.24).
“Research papers must be written in a difficult way (e.g., full of complicated equations and complicated figures) so that other readers think that the authors have deep understanding about their research topics.”
This mindset is indeed wrong. Writing a paper is similar to having a formal conversation between different researchers. If other researchers do not understand what you are talking about, this conversation is useless. Therefore, other researchers must not feel impressive about those “knowledge” that they do not understand. Maybe you can think in this way. If your friends, who also work in computer science research, say that they have developed a great algorithm but they cannot explain how great it is, you will not also think that this algorithm is great. Instead, you will have serious doubts about it. This situation is also the same when reviewers review your papers. If reviewers cannot understand many parts in your papers, they will also raise a lot of doubts about it. Obviously, with many doubts, they will ultimately give a “Reject” for that paper. Note that not many reviewers will give an “Accept” for a paper with a lot of doubts (especially for top-tier venues) because this is a responsibility for authors to present a paper so that readers can understand it easily. As a remark, for all research papers that (1) are submitted to top-tier venues, (2) are difficult to be understood, and (3) are reviewed by us, nearly all of them are rejected. We believe that other reviewers also have similar records.
Fig. 4.24
Illusion from students for writing complicated papers to confuse reviewers
Therefore, some smart students will be more considerate to the reviewers (see Fig. 4.25). They will assume that those reviewers do not have solid background for their research. Therefore, they will (1) draw some illustrative figures for explaining the core concepts (or ideas) and (2) make the paper more self-contained so that reviewers can easily understand how their techniques can advance the state of the art. Once reviewers are happy with the paper, it can significantly increase the chance for that paper to be accepted in a top-tier venue. Note that many papers (without any fancy technique) can also be accepted in top-tier venues. The main reason is that they are well-written (e.g., the story is interesting and the technique sounds correct, which can work in practice compared with existing techniques).
4.15 Overcomplicate Somethings
Many students may tend to overcomplicate somethings when they are writing research papers and discussing with others, e.g., supervisors (see Fig. 4.26). They have this kind of mindset mainly due to the following three reasons.
Fig. 4.26
Never overcomplicate somethings. Otherwise, it is a waste of time for both you and your supervisor
Show that they are “knowledgeable”/“smart”. Some students may think that discussing somethings in a complicated way can demonstrate that they are knowledgeable in their research topics or demonstrate that they are smart. We believe that they have this mindset because of some TV shows/movies, which depict that a “smart” scientist only talks about somethings that everyone (or at least a large portion of people) cannot understand. Therefore, they also need to follow this in order to be “smart”. To be honest, we think this is the most childish mindset. First, a researcher is knowledgeable in one research area because he/she has already provided solid contribution to it (e.g., publish a lot of top-tier papers or establish a great system that has been used by a lot of users) but it is not because he/she has made a lot of overcomplicated (a.k.a. useless) discussions for it. Second, a smart researcher only presents a complicated concept in an easy/intuitive way so that everyone in the same community (or even layman) can easily grasp the knowledge. This is very important because it can significantly reduce the communication time between different researchers.
Fig. 4.27
Weak researchers can overcomplicate somethings for their poor technical contribution, while strong researchers use the correct approach (i.e., think of new contributions/problem settings) for addressing this issue
Do not have enough technical contribution. Some students may discover that the technical contribution may not be enough. Therefore, they may try to add some unnecessary content (e.g., equations) into some sections in order to make the paper look “high class” (see Fig. 4.27). They may think that this is the only way to get rid of the comment of “not enough technical contribution” from the reviewer. However, we would like to point out that it is not very difficult for reviewers to determine the necessity of technical content (especially for those reviewers with solid reviewing experience) although reviewers may not be familiar with the research topics of the authors. Using the first author of this book as an example, he has reviewed one paper that has been submitted to a VLDB conference recently. He discovered that the paper has developed four algorithms for solving one problem. However, the first three algorithms are overcomplicated and not necessary. Therefore, he raised one question to the authors for asking why they need to make these unnecessary methods and also expressed the concern for the technical novelty/contribution of that paper. Ultimately, that paper has been rejected by the conference. Hence, the correct approach for solving the issue of “not enough contribution” is to think of new contribution or add new problem settings (see Fig. 4.27). There is no other approach to get rid of it.
Do not understand the concept well. If some students do not understand the concept well, it is normal for them to discuss it in a very difficult way. However, some students may still insist that they understand the concept very well even though they state that they cannot express it in an intuitive/easy way. Here, we need to emphasize that the concept must be easily explained by a student if he/she can understand it well enough (see Fig. 4.28). If it is still very complicated for the student to explain it, we can ensure that this student must not be able to understand this concept well enough. Therefore, what the student needs to do is to (1) take a look for that concept again and (2) figure out how to express it to others. Only those concepts that can be easily (and logically) expressed by the student should be regarded as understanding by that student.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
By reading this sentence, you need to be careful that it does not encourage you to avoid reading papers because of your laziness. You must not say this to your supervisor if you do not want to work.
