Publishing Papers BPR PhD 4US

By pjain      Published April 5, 2021, 1:36 a.m. in blog Education   

Results for PhD 4US

Advisor helps complete PhD in sane time

Critical Role of Advisors

If you do not have the equivalent of an advisor, it can get very hard to rotate over any work on your own with that timespan.

A direct way to get feedback on your research is basically what you will be doing in your PhD program although with some amount of initial hand-holding from your advisor.

Journals vs Conferences - peer reviewed

Publishing 101

Funding Impact - need Conferences AND Journal Publications

Some funding sources do not take conferences into consideration, so most of the top conferences are also published in special issues of top journals.

Publishing a Computer Science Paper is Hard

Journals

A direct way to get feedback on your research is by submitting to less selective second tier conferences or workshop and reviewer comments. After you build some foundation in that area, you can also reach out to other researchers, e.g., authors of papers you are using as references or professors working in that area.

On the other hand, the leading journal in my area, ACM Transactions on Database Systems, recently announced that if you get a paper accepted, you will get to present it at the next SIGMOD conference (albeit as a poster). I find it amusing that "quicker dissemination" is probably not an advantage to conference publication any more. Journals (in my area, at least) have done a good job of getting reviewing times down, plus papers are published on line as soon as accepted. Conferences have stretched out their reviewing times, by adding more steps such as reviewer bidding, author-comment periods, and on-line discussion periods.

Conferences

You could prove your capability by publishing to conferences, but you don't know which conferences are respectable and aren't just paper mills. Also, conference deadlines for the top conferences are few and far in between.

As in the other fields, the conferences are more prestigious than the journal. Extended versions of conference papers often appear in TOCS.

Conferences vs Journals

Review Time - Conferences < Journals

Because CS tends to move faster than other fields, and the emphasis is made on new ideas rather than careful vetting for quality. Conferences have a shorter review cycle with fixed deadlines, so a paper cannot be delayed by reviewers. This is so attractive that top conferences are more competitive than journals. However, competitiveness does not automatically translate into quality, as top conferences in some fields are very susceptible to fads (papers that are accepted because they cover interesting topic, even though the results may be mediocre).

We have this assumption that journals take long, detailed papers and do an extraordinarily careful job of reviewing and vetting—the way they do in CS. But in fact, many journals in other disciplines actually do very quick (ergo, can't be very detailed) reviewing, accept surprisingly short papers, or have fewer reviewers, or more “interference” from editors. So we have a bit of peril in CS of automatically granting too much dignity to journals in other disciplines, as if they were like journals in CS.

In general, I think many of them are somewhere in-between, but good conferences (not necessarily the same as “top” conferences)—i.e., conscientious ones—do a very thorough job of reviewing within the constraints of a conference review cycle (though processes like response periods, two-stage submissions, etc. are all even further narrowing that gap).

Reviews more indepth in Journals vs Conferences

Conference reviews have gotten in 2014 more detailed and involved. Early in my career (several decades ago), all you got back from a conference submission was a "yes" or "no". Over the intervening years, almost all conferences I submit to (or review for) have gone to a multi-section review form with strengths, weaknesses, detailed comments, etc. Some conferences in my area have recently put in revision cycles, where some authors are allowed to revise and resubmit, if reviewers think the paper can be fixed in a short time. One of the major conferences in my area (Very Large Data Bases) has gone to year-around submission. Papers are published on line periodically, and are then presented at the next meeting of the conference. Multi-stage review with papers appearing at intervals throughout the year is looking a lot like a journal. On the other hand, the leading journal in my area, ACM Transactions on Database Systems, recently announced that if you get a paper accepted, you will get to present it at the next SIGMOD conference (albeit as a poster). I find it amusing that "quicker dissemination" is probably not an advantage to conference publication any more. Journals (in my area, at least) have done a good job of getting reviewing times down, plus papers are published on line as soon as accepted. Conferences have stretched out their reviewing times, by adding more steps such as reviewer bidding, author-comment periods, and on-line discussion periods.

It is not that the prestige of journal papers is lower, rather large number of high quality research papers are also submitted to top-tier conferences making them equally and more prestigious. The reason why computer scientists prefer submitting their work to conferences over journals has to do with long and slow review cycles of journals. As opposed to pure sciences and mathematics, for journals in computer science finding qualified peers to review a paper in a particular subarea of computer science throughout a year is harder. This leads to long and slow review cycles where submission to publication time varies between 4 months (if you are extremely lucky) to 2 years. For a fast moving discipline like computer science, the methods and findings become very outdated in that span of time. On the other hand all good computer science conferences I know of have very focused and rigorous review process with most having a rebuttal phase too. Yet, from submission to publication, time taken is from 2 to 4 months, making top-tier conferences the primary choice for publishing high-quality research. SRC

BPR 1. CHOOSE CONFERENCE/Journal for FIT

  • While conferences are different, there's quite a lot of overlap. For some papers, it's immediately obvious that it came from one or the other, but many could fit in either.

BPR 2. Best conferences are considered more prestigious than the best journals.

However the line between conferences and journals in CS is rapidly blurring.

Conference searches

Microsoft Academic Search has a nice interface for showing top CS conferences/journals.

Here are a few links for some disciplines: - Information Retrieval - http://academic.research.microsoft.com/RankList?entitytype=3&topDomainID=2&subDomainID=8 - Machine Learning - http://academic.research.microsoft.com/RankList?entitytype=3&topDomainID=2&subDomainID=6 - Data Mining - http://academic.research.microsoft.com/RankList?entitytype=3&topDomainID=2&subDomainID=7 - Computer Vision - http://academic.research.microsoft.com/RankList?entitytype=3&topDomainID=2&subDomainID=11 - World Wide Web - http://academic.research.microsoft.com/RankList?entitytype=3&topDomainID=2&subDomainID=15&last=5

No broad CS Conferences

Unlike, say, the Annals or Acta in mathematics or Nature or Science in biology and some other fields, there are no venues in computer science that cover a broad swathe of subject matter and that researchers across the discipline aspire to publish in.

So it may take several months to determine if your paper has been accepted.

In computer science, the prestigious venues for publication are not journals -- they are conferences. Each sub-area has their own top conferences

There's not a dominant venue that covers the whole field. In many sub-fields, the ACM conference in whatever your sub-field happens to be is the best. There are also very good conferences in certain sub-fields by IEEE, USENIX, and a couple of others.

Academic research Builds on Previous work

As others have said, the claim is that CS moves so fast that journal publication takes too long. But you may see CS researchers "reinviting the wheel," addressing problems that have been solved previously in other fields, notably statistics.

101: Subfields change rapidly, go obsolete

As others have said, the claim is that CS moves so fast that journal publication takes too long.

Academic research is supposed to build on previous work, and I would argue that CS doesn't really do that. It's not that advances are made rapidly within a subfield, but rather that the subfields themselves don't last long, maybe a year or two. A few years ago, for instance, Sensor Networks were big, then suddenly it became a nontopic, as far as I can tell. A few years ago, Data Mining was big. Then it became Big Data. Then it became Data Science. Granted, there has been some "institutional memory" during that transition, but overall I just don't see the degree of building up of knowledge that is supposed to be the essence of research. - SRC

Conferences and Journals by Area

Machine learning

  1. NIPS.cc - big general conference

  2. ICML.cc - big general conference

  3. COLT is up there for learning theory. http://orfe.princeton.edu/conferences/colt2014/

Computer architecture

  1. ISCA

  2. MICRO (Focus on Microarchitecture)

  3. HPCA

  4. ASPLOS (Interdisciplinary : Computer Arch + Programming Languages + OS)

Programming Languages

While POPL and PLDI are the most prestigious, there are also some more specialized conferences somewhat "less prestigious" than POPL or PLDI.

  1. Principles of Programming Languages (POPL)--more theoretically inclined

  2. Programming Language Design and Implementation (PLDI)--more implementation oriented than POPL

  3. ASPLOS (Multidisciplinary : Computer Arch + Programming Languages + OS)

  4. PACT (Multidisciplinary : Architecture + Compiler)

  5. PPoPP (Focus on Parallel Computing)

  6. Object-Oriented Programming, Systems, Languages and Applications (OOPSLA)--about OOP, with some focus on object-oriented languages and rather less theory.

  7. European Conference on Object-Oriented Programming--like OOPSLA but more European? I've actually never seen anything from this one, but apparently it's big.

  8. International Conference on Functional Programming (ICFP)--focuses on implementation and design of functional programming languages, with a fair amount of theory.

  9. SC,ICS (Focus on Supercomputing)

--- Theoretical CS - annual

  • FOCS, the IEEE Symposium on Foundations of Computer Science, around October;
  • STOC, the ACM Symposium on Theory of Computation, around the beginning of June;
  • SODA, the ACM-SIAM Symposium on Discrete Algorithms, around early January.
  • CRYPTO cryptologists may rate it at the same level or higher for their purposes, etc.

Some people with more of a complexity than an algorithms bent will rate only FOCS and STOC as top conferences and put SODA in the next tier

Databases

  1. Very Large Data Bases Conference 2014 has gone to year-around submission. Papers are published on line periodically, and are then presented at the next meeting of the conference. Multi-stage review with papers appearing at intervals throughout the year is looking a lot like a journal.

  2. ACM Transactions on Database Systems - Journal. In 2014 announced that if you get a paper accepted, you will get to present it at the next SIGMOD conference (albeit as a poster). I find it amusing that "quicker dissemination" is probably not an advantage to conference publication any more. Journals (in my area, at least) have done a good job of getting reviewing times down, plus papers are published on line as soon as accepted. Conferences have stretched out their reviewing times, by adding more steps such as reviewer bidding, author-comment periods, and on-line discussion periods.

Networking

Operating systems

  1. Symposium on Operating Systems Principles (SOSP). Alternates each year.

  2. Symposium on Operating Systems Design and Implementation (OSDI). Alternates each year.

3, Nineteenth International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) and EuroSys are also considered good conferences to publish in.

  1. The top journal for operating systems is the ACM Transactions on Computer Systems. As in the other fields, the conferences are more prestigious than the journal. Extended versions of conference papers often appear in TOCS.

Graphics

The best conference for computer graphics is ACM SIGGRAPH, which also gets published in ACM Transactions on Graphics (TOG) (recently, another issue has been assigned to SIGGRAPH Asia).

For visualization, IEEE Visualization is the leading conference (published as a special issue of IEEE Transactions on Visualization and Computer Graphics (TVCG)). Regular issues of both TOG and TVCG are also considered prestigious.

For Human-Computer Interaction, CHI is the leading conference.

Biomedical CS

For biomedical applications of CS (where the line between CS/EE/ME get a little blurry).

  • IEEE Trans. on Medical Imaging

  • IEEE Trans. on Biomedical Engineering are good venues

  • MICCAI is another prestigious biomedical conference.

Research Process BPR

No Entry Barriers in CS

Computer Science is nice in that you require no expensive specialized hardware in order to work on research (as always, with exceptions).

Get into a Good PhD Program

If you can demonstrate that you are capable of doing independent research, you should be likely to get into any PhD program of your choice.

Do a Literature Broad Sweep - Follow your Interests

You can get started by reading as many highly cited papers as possible from top tier conferences in your area of interest; you can find rankings on multiple websites.

After a point you would see a pattern of problems that multiple people are working on and are if interest to the community.

Publishing in a Job

If you like the work you are doing at job, just explore and dive deep into it Read some latest journals and top tier conference publications in your area to try and synchronize your work and interest to common goal. Moreover if your company is into research, you always have ways to publish into journals and conferences with some effort. Also most research based organizations do have some good contacts into academics, try and leverage them from your current organization.

Choosing a Topic

However, things like figuring out the precise research topic (image processing is WAY too broad) and learning what is cutting-edge in your field as well as the major outstanding problems takes a lot of time.

Do a Literature Indepth in your Topic

Novel Data Sets or experiments may give you an edge

Try to find public domain datasets and see if you can reproduce or improve on those results.


0 comments

There are no comments yet

Add new comment

Similar posts

K8 Tutoring Young Children ST4US

Space Companies and Investing

Space Science Industry Review, Applications, Keys and Uses

Indian Auto and EVs Startups