Jump to content

Guidelines for Presentations/Proposal Talk

From SDQ-Wiki
This page is a translated version of the page Vortragshinweise/Proposalvortrag and the translation is 100% complete.

Proposal Talk

Goals

  • Convey to the audience which problem you will address in your thesis, why solving the problem is important, which approach you are pursuing, and how your approach is to be evaluated.
    • It must be clear that you approach your thesis with scientific methodology.
    • You must make clear that the problem is relevant and still unsolved or insufficiently solved.
    • You must compare your approach with similar work from the literature (related work).
    • You must show how you intend to demonstrate that your approach actually delivers the promised improvements (form of evaluation, e.g., case study, experiment, formal proof).
  • The speaker uses the talk to obtain feedback on the planned work from the researchers. They can
    • point to further related work
    • assess your methodology
    • share experiences with the tools you intend to use
    • give hints on how to improve the approach
  • The talk should therefore be designed so that it enables the researchers to assess the points mentioned (e.g., by listing related work, tools used, etc.)

Mandatory Content

  • Organizational matters
    • Who is the co-supervisor? (If this has not been determined yet, you should ask about it at the latest during the talk)
    • When are code reviews planned?
  • For implementation work: We want to reuse as much existing work and implementation as possible. Therefore, also address
    • which existing implementations of the chair (PCM, Vitruvius, etc.) you want to build upon
    • where you want to extend them

General Structure

  • About 6–9 slides
  • No outline slide with a table of contents (an overview is not necessary for a talk of only 12 minutes)
  • However, a progress bar/tab index on each slide is quite useful for orientation
  • No lengthy introduction: after 2–3 slides it must be clear what you want to do in your thesis
  • First present what you want to do, and give details later on why it is important and how you want to do it
  • If related work only serves to delineate your work and you do not build on it, present it at the end of the talk so as not to interrupt the presentation of the approach unnecessarily.

Possible Structure (Not Mandatory)

  • Slide 1: Motivation and illustration of the problem (and, if applicable, your approach), graphically if possible (on one slide!)
  • Slide 2: Instead of an outline slide, show a PIBA (four bullet points).
    • Problem, Idea, Benefit, Actions (PIBA)
      • Problem: What is the problem? (understandable to any computer scientist)
      • Idea: What is your approach?
      • Benefit: What is the benefit of your approach? (also compared to other approaches)
      • Actions: Which steps do you want to take?
    • Depending on the topic, three aspects may also suffice (e.g., P-I-B); PIBA is not a strict requirement
  • Slide 3: Foundations: What are the most important foundations of your subject area?
  • Slide 4: Related work: What has already been done in the research area? (alternatively after the evaluation, if only for delineation)
  • Slide 5: Concrete, simple example of your approach (illustrated)
  • Slide 6: Implementation of your approach
  • Slide 7: Planned evaluation
  • Slide 8: Schedule
  • Slide 9: Summary; simply show the PIBA slide (slide 2) again.
    • The slide does not have to be an exact copy of the PIBA slide; it can also omit aspects and include additional points such as evaluation or future work

Motivation

If you need to motivate, show the problem concretely using an easily understandable and usually simple case. Examples:

  • For prediction methods: Show predictions that you consider too imprecise. Show the targeted improvement already here.
  • For tool implementations: Show how much effort it is to perform a central task without your tool. Give an initial impression of the complexity of automating this task.
  • For experiments/empirical studies: Show the research question of your study and why an answer to it is a gain in knowledge. And above all, for whom the findings are important.

Foundations

  • Introduce the most important technical terms from your subject area (at least those from the title of the thesis!)
  • Explain fundamental concepts/models (e.g., the Palladio Component Model)

Related Work

  • Use names and years of the publications (e.g., [Balsamo 2004]), and keep the correct references (title, authors, form of publication, etc.) ready on a backup slide.
  • For each cited work: the title, or in one bullet point what is done there
  • First 2–3 foundational works, standard works (books, important articles),
  • then the 3–4 most relevant related works you have found
  • Delineate your work concretely from the latter. What is possible with your approach that is not possible with these approaches? How much do you want to improve individual parts of the related approaches?
  • However, do not condemn related work. Be aware that your approach often improves only one detail, and that your approach usually does not have all the strengths of the related work. Nobody believes that your approach solves all problems.
  • Do not forget industrial solutions (that is, do not only search Google Scholar and the computer science library)! Often there are already commercial approaches. Try to assess them as well as the publicly available data allows.

Example

  • It should demonstrate as simply and understandably as possible what happens when your approach is applied or your product is used.
  • Build it up step by step, possibly with animation.
  • Images are easier to understand than text.

Implementation of Your Approach

  • Which steps do you want to take?
  • Demonstrate the iterative structure of the work with work packages (including optional ones).
  • Couple work packages with the schedule. (A Gantt chart is usually too cluttered. Limit yourself roughly to 3–4 work packages. Writing the thesis is not a work package, but part of the schedule. Programs for creating Gantt charts include Microsoft Project and Microsoft Visio.) The schedule should also include the planned code reviews.
  • Demonstrate a systematic and goal-oriented procedure.
  • Risk management (What risks are there? How do you plan to react to them, if necessary?)
  • Which tools are to be used? (Name them concretely)
  • Be as concrete as possible on this slide. Not "I will create a model transformation for concurrency patterns", but "I will use the eXtend language from the OAW framework to implement a model-to-model transformation. The design patterns Replication and Rendezvous contained in the source model are mapped in the target model to parallelism operators of the stochastic process algebra Capra." Not "I will solve the problem with LLMs", but "I will use prompts such as <example prompt> with this additional context information."

Planned Evaluation

  • Form (case study, controlled experiment, formal proof)
  • Case example (object of study during the evaluation)
  • Goal, Question, Metric (What goal does the evaluation pursue? Which questions are derived from the goal? Which metrics must be collected to answer the questions?)
  • Hypotheses (What result do you expect/hope for? What are the expected answers to the questions?)
  • The evaluation must be designed so that a computer scientist not familiar with the subject area can judge whether your work was a success or not.

Schedule

  • Show a shortened schedule that reveals the basic structure of your procedure.
  • Make clear how you ensure incrementality in your work. A waterfall-like process is usually questioned, as the risk of failing in an early phase and producing no results is too high.
  • Explicitly show time for buffers, interim evaluations, or milestones.
  • Activities can quite well overlap in the schedule. For example, you can start writing early while still working on other work packages.

Summary

  • Feel free to show the PIBA slide again here as a basis for discussion.
  • The slide does not have to be an exact copy of the PIBA slide; it can also omit aspects and include additional points such as evaluation or future work
  • Still summarize the problem statement and your planned work once more in 2–3 sentences, and do not end merely with the words "Here is the PIBA slide once more".