21
Research Methods and Techniques Lecture 2 Paper Review 2 © 2004, J S Sventek, University of Glasgow

Research Methods and Techniques Lecture 2 Paper Review 2 © 2004, J S Sventek, University of Glasgow

Embed Size (px)

Citation preview

Research Methods and Techniques

Lecture 2

Paper Review 2

© 2004, J S Sventek, University of Glasgow

7 October 2004 RMaT/Paper Review 2 2

Reminders/Notices Web site:

www.dcs.gla.ac.uk/~joe/Teaching/RMaT.html Course director: Prof J S Sventek

[email protected] Assignment due next Thursday, 14 Oct 2004

One page summary of paper 2 Lectures rescheduled

Old New Room Other

21 Oct 18 Oct F171 Assignment due on the 21st to be handed into MSc office (S143) by 16:00 BST

4 Nov 1 Nov F171

7 October 2004 RMaT/Paper Review 2 3

Review Template Briefly summarize the paper in 3-5 sentences. What problem does this contribution solve? Is it an

important problem? What are the claimed contributions of the paper? Upon what (and whose) previous work has this

research been based? What methodology has been used? Is it appropriate? What conclusions are drawn from the results? Are

they correct? Has the research been performed correctly? Is the presentation satisfactory? Is the paper appropriate for the target venue?

7 October 2004 RMaT/Paper Review 2 4

Strategies for Reading Papers

First skim the whole paper once to get a feeling for what it is about. Skip anything dense or technical. Glance at the figures and tables.

Read through the whole paper, but don't let yourself get stuck. If you have questions, don't struggle to figure them out; instead, jot a note in the margin and keep going. Put stars next to important points and/or underline key ideas. Don't get carried away with the underlining, or it won't be helpful when you want to look back over the paper.

Go back and work through the details of any (important) equations, proofs, algorithms, etc. that you skipped over.

Now fill in the review template with the information you have gleaned from your reading.

(borrowed from http://www.cs.umbc.edu/~mariedj/691b/paper-summary-guidelines.html)

7 October 2004 RMaT/Paper Review 2 5

Strategies for Reading Papers (2)

Even for papers for which you don't need to turn in a research summary, always write a 1-page summary of the paper in your research journal. When reviewing a paper, or reading it seriously, try to make thorough enough marginal notes that you can extract a summary/review directly from these notes. When reviewing a paper for a conference or a journal, it is also helpful to make summary notes ("great idea but poorly executed," "thorough results but uninteresting work") on the first page.

You should maintain a research journal! This should be a bound composition book, not a bunch of scraps of paper or a ring binder. Researchers often look back at their recorded reviews when tackling a new problem.

(borrowed from http://www.cs.umbc.edu/~mariedj/691b/paper-summary-guidelines.html)

7 October 2004 RMaT/Paper Review 2 6

Review of Sample Paper

Paper Title A Non-invasive Platform Supporting Distributed, Real-Time, Multimedia Collaboration

Author[s] M C Hao and J S Sventek

Summary This paper describes a platform that supports use of existing multimedia applications in a distributed, collaborative environment. It does this without requiring access to or modification of the applications themselves by providing a proxy capability for intercepting window system events and replaying them to multiple, synchronized copies of the application being used. It shows significant performance gains relative to centralized and/or invasive collaboration platforms.

Problem

Solved

There are many industries today that rely upon the use of “sophisticated, computer-aided design tools” to produce their products. Often, it is desirable that many employees collaborate over the design and use of the designs resulting from use of such CAD tools.

7 October 2004 RMaT/Paper Review 2 7

Review of Sample Paper (2)Problem

Solved

(cont)

Previous solutions to this problem are: 1) a single version of the CAD tool is executed, with a pseudo-window-server interposed between the users and the application; this solution generates significant network traffic due to the continual transmission of graphics primitives and bitmaps to the collaborating workstations; and 2) the application to be shared is modified so that the application instances explicitly collaborate; multiple copies of the application are started up, one per user.

The invasive approach described above is usually not an option for most applications that can be shared. The pseudo-window-server approach, while being non-invasive, requires significant bandwidth, especially if the application is communicating large bitmaps. This system also does not work if the applications use Direct Hardware Access to the bitmap system, as opposed to performing all output through the window server.

The system described in this paper provides a different level of indirection that is non-invasive, yet does not suffer from the large bandwidth requirements and can work with DHA graphics.

7 October 2004 RMaT/Paper Review 2 8

Review of Sample Paper (3)Claimed

Contributions

1. non-invasive collaboration platform

2. low bandwidth requirements

3. works with applications that use DHA graphics

Directly-

Related

work

VConf system of Lantz and Lauers – similar approach integrated with the window server, did not implement floor control; propose to make applications collaboration-aware, thus becoming invasiveSharedX, a centralized, non-invasive system; will not support applications using DHA, when used with CAD tools the bandwidth requirements are hugeMMConf, an exemplar of the invasive type of platform, only requires that an application be relinked to a substitute library; at the time of this paper, DLL’s were in their infancy, so this approach would have to be revisited now.

Other

Related

Work

NetMeeting from Microsoft, centralized, based upon the Windows API, so Windows-specific. Since it is a commercial product, have less visibility in the design and architecture of the system; cannot work with DHA, bandwidth requirements will be very large.

7 October 2004 RMaT/Paper Review 2 9

Review of Sample Paper (4)Methodology Design of indirection mechanism, engineering of the mechanism,

experimental trials with typical CAD tools used in the automotive and aerospace industry, beta test of research prototype in industrial setting, brief summary of experience in both types of trials

Conclusions It is possible to build a non-invasive, distributed collaboration framework that will work in bandwidth-limited environments and with applications that use DHA graphics.

Do Differently?

None stated.

What did you learn?

Nothing, I wrote the paper.

7 October 2004 RMaT/Paper Review 2 10

Conversion to a 1-page summary

The Summary field in the review becomes the abstract for your summary.

The Problem Solved field usually describes the problem solved and how it was solved.

The remaining narrative of the 1-page summary is derived from the Problem Solved, Claimed Contributions, Related Work, Methodology and Conclusions sections.

If the author[s] specified anything that should be done differently, you should try to weave something about that into your narrative.

7 October 2004 RMaT/Paper Review 2 11

1-page Summary of Sample Paper

Abstract: This paper describes a platform that supports use of existing multimedia applications in a distributed, collaborative environment. It does this without requiring access to or modification of the applications themselves by providing a proxy capability for intercepting window system events and replaying them to multiple, synchronized copies of the application being used. It shows significant performance gains relative to centralized and/or invasive collaboration platforms.

There are many industries today that rely upon the use of “sophisticated, computer-aided design tools” to produce their products. Often, it is desirable that many employees collaborate over the design and use of the designs resulting from use of such CAD tools..

Two types of solutions to this problem have been discussed in the literature prior to this paper. In the first, a single version of a particular CAD tool is executed, with a pseudo-window-server interposed between the users and the application; this solution generates significant network traffic due to the continual transmission of graphics primitives and bitmaps to the collaborating workstations. In the second, the application to be shared is modified so that the application instances explicitly collaborate; multiple copies of the application are started up, one per user.

Title: A Non-invasive Platform Supporting Distributed, Real-Time, Multimedia Collaboration

Name: J S Sventek

Author[s]: M C Hao and J S Sventek Matric number: 1234567

7 October 2004 RMaT/Paper Review 2 12

1-page Summary of Sample Paper (2)The invasive approach is usually not an option for most applications that can be shared, as the source is not available. The pseudo-window-server approach, while being non-invasive, requires significant bandwidth, especially if the application is communicating large bitmaps. Such an approach also does not work if the applications use Direct Hardware Access to the display subsystem, as opposed to performing all output through the window server.

The system described in this paper provides a different level of indirection that is non-invasive, yet does not suffer from the large bandwidth requirements and can work with DHA graphics. It does so by interposing a proxy process between the window system and each user’s copy of the shared application; this proxy captures all window system events from the user (mouse clicks, keystrokes); it distributes these received events to all of the other proxies in this collaborative session; each proxy then proceeds to deliver these events to its application for execution.

Since multiple users could attempt to make changes to a design at the same time, the platform introduces a conservative form of floor control, such that only one individual can be making changes at a time; the floor is explicitly requested and released; a request for the floor when someone else possesses the floor causes the request to be queued until the floor is released.

7 October 2004 RMaT/Paper Review 2 13

1-page Summary of Sample Paper (3)The authors relate their work directly to the VConf system, which took a similar approach by integrating this support directly in the V system window server; unfortunately, this system did not implement floor control, and proposed that applications become collaboration-aware, thus making that system invasive. They also compared against SharedX, which is a centralized, non-invasive approach to the problem; SharedX exhibits large bandwidth requirements when used with CAD tools, as well as being unable to enable DHA applications to collaborate. Finally, they compare against the MMConf system, which is an exemplar of an invasive platform; it is probably the simplest form of invasive system, in that it only required that applications be relinked against a collaboration-aware library. They also discussed NetMeeting from Microsoft, which is a proprietary, centralized, Windows-specific collaboration platform; it has the same characteristics as SharedX.

The paper uses a traditional methodology of designing the new indirection mechanism, implementing it, running experimental trials with typical CAD tools used in the target industries (automotive and aerospace), and beta testing of the research prototype in industrial settings. They briefly summarize the experience in both types of trials.

The paper gives an existence proof that it is possible to build a non-invasive, distributed collaboration framework that will work in bandwidth-limited networking environments and with applications that use DHA graphics.

7 October 2004 RMaT/Paper Review 2 14

Giving Due Credit to Others Whenever one writes technical prose, especially

when reviewing the work of others, it is important to give due credit to the author

Plagiarism is using the work of others and misrepresenting it as your own.

This includes: Cut & paste from a reading assignment Cut & paste from the web Cut & paste from books, other papers, etc. Cut & paste from ANYTHING that is not your own! Changing wording of a sentence but keeping the ideas Summary which does not include proper references

7 October 2004 RMaT/Paper Review 2 15

Glasgow Plagiarism Policy Plagiarism is defined as the submission or presentation of work,

in any form, which not one’s own, without acknowledgement of the sources. Special cases of plagiarism can also arise from one student copying another student’s work or from inappropriate collaboration.

The incorporation of material without formal and proper acknowledgement (even with no deliberate intent to cheat) can constitute plagiarism. Work may be considered to be plagiarised if it consists of: a direct quotation a close paraphrase an unacknowledged summary of a source direct copying or transcription

7 October 2004 RMaT/Paper Review 2 16

Glasgow Plagiarism Policy (2) With regard to essays, reports and dissertations, the rule

is: if information or ideas are obtained from any source, that source must be acknowledged according to the appropriate convention in that discipline; and any direct quotation must be placed in quotation marks and the source cited immediately. Any failure to acknowledge adequately or to cite properly other sources in submitted work is plagiarism. …

7 October 2004 RMaT/Paper Review 2 17

Example: Repeating Another’s Words Without Acknowldegement

Original quote“To capture the various interaction styles of publish/subscribe, we propose an abstraction called Distributed Asynchronous Collection (DAC). A DAC differs from a conventional collection by its distributed nature and the way objects interact with it: besides representing a collection of objects (set, bag, queue, etc.), a DAC can be viewed as a publish/subscribe engine of its own. In fact, when querying a DAC for objects, the client expresses its interest in such objects. In other words, the invocation of an operation on a DAC expresses the notion of future notifications and can be viewed as a subscription.”

Plagiarized VersionA Distributed Asynchronous Collection (DAC) represents a collection of objects; a particular distributed asynchronous collection can be viewed as a particular type of publish/subscribe engine. Invocation of an operation on a DAC can be viewed as a subscription, since it expresses interest in future notifications.

Correctly Documented VersionEugster et al [Eug00] describe a type of dynamic collection, a Distributed Asynchronous Collection, where an operation on an instance of such a collection actually represents a statement of interest in future notifications in the publish/subscribe engine that the collection represents.

[Eug00] P Eugster, R Guerraoui, J S Sventek, “Distributed Asynchronous Collections:Abstractions for Publish/Subscribe Interaction”, Proceedings of ECOOP 2000, LectureNotes in Computer Science, Vol 1850, pp. 252-276, Springer-Verlag, 2000.

7 October 2004 RMaT/Paper Review 2 18

This is how it is …

Lecturers can instinctively sense plagiarism They are likely to know or to have seen the material

from which you are quoting They often conduct extensive searches to check for

plagiarism, especially in dissertations Therefore, read and understand the material

(whether it is assigned, or related to your literature search/project plan), but be sure to always acknowledge the origin of the material.

7 October 2004 RMaT/Paper Review 2 19

One method for avoiding plagiarism

Read the assigned material Read over any additional material that you

think you need (e.g. follow up on directly-related references)

Put everything away Fill in the template (review/in-depth review) –

this forces you to use your own words Add proper citations and references to the

materials that you have used.

7 October 2004 RMaT/Paper Review 2 20

Strategies for Reading Papers (3)

If you're going to present the paper at a class, or if it's very relevant to your own research, you may need to spend more time with the paper to fully understand it. You may also need to look up citations of work (by the authors or by others) on which this paper builds. It's important, though, to learn which papers are critically important and which are less so -- it would be entirely possible to spend the next year (or two or three or four) doing nothing but reading, and still not have read all of the papers that you "ought" to. Spend your time wisely.

Summaries that go into this additional level of detail are what we call “in-depth reviews”.

(borrowed from http://www.cs.umbc.edu/~mariedj/691b/paper-summary-guidelines.html)

7 October 2004 RMaT/Paper Review 2 21

In-depth Reviews 3-5 pages long If done correctly, someone else that reads your in-depth review

will completely understand the paper without having to read the paper

Requires that you review the directly-related references, so that you can summarize that work in your own words, not just rely upon the 1-2 sentences that the author[s] of the paper you are reviewing have provided to differentiate the current work from the related work.

When you review papers while developing your project proposal, you need to generate summary reviews of all work to which your project is directly related

You will then aggregate the information from these reviews into the “Related Work” section of your project dissertation.