Application Process for Thesis

Requirements

Since we get many requests and have limited capacity, we require that you have taken relevant courses with us (such as Introduction to Deep Learning) before starting your thesis with us. Alternatively, you need to provide a transcript of records as well as a recommendation by another (FIN) Professor vouching for your skills in the relevant area(s). This way, we can make sure you bring the prerequisites to successfully complete challenging work in the field of Deep Learning independently.

Step 1: Familiarize yourself with available topics and match them with your research interest

The ongoing open theses topics are published here. You are also welcome to come to us with your own ideas for projects and theses. This could also be in cooperation with another research group or a company.

Step 2: Get in contact

You found something that sparked your interest? Please fill out the application form.
Your submitted information will be solely used for the application process. It will be treated confidentially and deleted latest 6 month after the end of the application process.

Step 3: Short discussion

Based on our internal capacity, we will invite you to a short discussion to get to know each other, get an understanding of the topic and discuss further steps. After this discussion, both you and your potential supervisor can decide if you want to go forward with it and get started.

Step 4: Start the thesis/ project

See Thesis Process below for more details.

 


 

Thesis Process

Disclaimer:

These are just hints on how we usually approach theses. Please refer to the official requirements of your course (especially regarding submission, formal requirements and defense).

 

Step 1: Getting into the topic

Depending on your previous knowledge, the actual thesis work starts with an introductory period, which covers the following tasks:

  • Familiarizing with the literature in the chosen topic
  • Familiarizing with the system environment, in which the implementation takes place
  • Narrowing down the subject
  • Understanding the problem that you are addressing properly

The result of this phase should be a short description that precisely defines the exact setting of the task and specific goals. Ideally, this description should already include a comparative presentation of possible solutions considering the literature, a short description of the chosen solution, and a working schedule. This procedure is supposed to avoid problems in understanding when defining the thesis and serious false estimations in the expenditure of work and with these overdrawn working periods for theses. Therefore, it is also true for external theses, at which the institute is the second expert.

The results of this phase will be presented and discussed in your kick-off presentation in our AI Lab Colloquium.

Step 2: Working on the thesis

In the actual thesis, the work will be done according to the presented working schedule. This means that first of all the chosen solution will be worked out in detail. Then, the system will be implemented. After that, the evaluating analysis and documentation follow. Finally, the results will be worked out in a written form. When grading the work, the quality of the implementation and formulation is as important as independent work in narrowing down the subject, literature research and finding the solution. It is also important to critically evaluate the design alternatives from the point of view of computer science and if applicable of the application field (this includes later self-criticism of the chosen alternative).

For practical work, the written formulation should include the following items. This list doesn’t claim to be complete.

  • Introduction
    • Introduction into the topic of the thesis
    • Narrowing down the task and problem
    • Answers to the questions: Why is the thesis useful? What is the goal of the work? What is the problem you are solving?
  • Determine and discussion of the starting point
    • What relevant methods/techniques are already developed in this research area (literature research)?
    • Which conditions (for implementation: hardware and software among others) exist?
  • Development of a (formal) solution model
    • Critical evaluation of the literature: What can be used?
    • Starting from the previous item, a concept for solving the task has to be developed
    • Justification of the solution concept: Comparision to alternatives! This section must not include implementation details.
  • Implementation of the solution concept
    • Selection of tools (e.g. programming language)
    • Documentation of important design decisions
    • Critical evaluation of the implementation (test runs with measured run times, case study if possible, gaps in the implementation, useful future additions,…)
  • Conclusion
    • Summary: What are the important results?
    • Future work: What can/must be done?

The formulation is not supposed to be a documentation of the developed programs. It is intended to show the ability to work with or develop methods of computer science independently. Used results out of the literature have to be marked thoroughly to separate them from the own work. Programming language expressions like “if a=b then s1” must be separated from common speech and other notations (like mathematical expressions) to avoid misunderstanding.

In addition, the thesis should include the following items:

  • Title Page
  • Abstract (one page at most)
  • Table of Contents
  • Table of figures and table of tables if applicable
  • Table of abbreviations (consists of all used abbreviations and symbols)
  • Appendix if necessary
  • Bibliography
  • Declaration of independent work (signed), e.g.: “With this, I declare that I have written this paper on my own, distinguished citations, and used no other than the named sources and aids.”

We recommend using the Overleaf Latex Template created by Alexander Dockhorn.

Phase 3: Defending your thesis

The final presentation takes place after handing in the thesis. It will take one hour and consists of

  • a 25-minute presentation of the graduate (20-minute presentation for bachelors),
  • the reading of the expertises (not for a “Studienarbeit”) and
  • a discussion between the graduate and the other participants of the presentation.

The presentation should include the following items:

  • Short introduction to the problem,
  • Short overview about relevant literature or rather methods and results from other projects that contributed to the thesis,
  • Presentation of the solution concept and implementation decisions (no program code, but important design decisions),
  • Presentation of the results (short demonstration of the implemented system, if possible) and
  • Summary, critical evaluation of the own work, and suggestions for future work.

 


 

Theses Examples

Here are some examples from past successful theses with the Artificial Intelligence Lab.

Master Theses

Speech Recognition and Generation using Capsule Networks
Towards Safe Handling of Sensor Data in Cooperative Systems

Last Modification: 30.11.2021 - Contact Person: Michael Preuss