Doing a Master Thesis on Applied Machine Learning

Applied machine learning (ML) transforms data into meaningful insights across domains—from healthcare and finance to climate science and beyond. If you're interested in exploring a research question or application using ML, consider having me as your thesis supervisor.

You can find a list of master theses currently supervised by me here.

What can you expect from me as your supervisor?

As your supervisor, you can expect me to:

• Help you clearly define your ML problem.
• Advise you on suitable ML methods, tools, and resources.
• Provide regular feedback during our group meetings.
• Guide you through thesis writing and evaluation.

My Research Interests

My expertise includes:

• Networked data analysis: Leveraging relational structures in data (recent example).
• Explainable ML: Using information theory to measure and improve interpretability (example paper).
• Applied ML: Exploring novel and impactful applications of ML techniques.

I'm also open to supervising projects outside these areas if the topic is exciting and well-defined.

Getting Started

To start your thesis:

1. Formulate your ML problem clearly (watch this video).
2. Choose suitable ML models you're comfortable implementing (e.g., linear regression, neural networks).
3. Identify data sources, features, and evaluation criteria (e.g., test accuracy, computational efficiency).

You can find detailed guidance in Chapter 2 of my textbook.

Additionally, I've prepared lecture videos on these topics.

Practical Workflow

Your thesis typically involves iterative steps of:

• Data collection and preprocessing using pandas.
• Model selection, training, and evaluation using scikit-learn.

To validate your methods, consider reviewing peer-grading guidelines used in:

• Machine Learning course
• Federated Learning course

Thesis Manuscript Preparation

When preparing your thesis, ensure:

• The ML problem is precisely defined (data, features, labels).
  (Check definitions using the Aalto Dictionary of ML).
• Your results are thoroughly analyzed and clearly presented.
• Appropriate baselines or benchmarks are used (e.g., Kaggle competitions).
• Chapters and sections are clearly structured, with introductory paragraphs explaining content and connections.
• Equations, figures, and tables are consistently referenced using LaTeX commands (\ref{}, \eqref{}).
• New methods are presented as pseudocode (see examples).
• Figures are clear, labeled, and have informative captions (guidelines).
• References are formatted according to the IEEE guidelines.

For guidance on creating effective and visually appealing figures, consider consulting Edward Tufte's classic book "The Visual Display of Quantitative Information".

Iterative Writing Process

Expect multiple iterations on your manuscript:

• Write your first draft quickly to capture your main ideas.
• Regularly incorporate feedback from peers, group meetings, or even LLM tools.
• Consider writing non-linearly, starting from results or discussion and working backwards to the introduction.

Final Thesis Checklist

Before submitting your thesis, verify:

• Clear problem formulation
• Methods clearly described, including pseudocode
• Clear figures with labeled axes and informative captions
• All equations, tables, and figures referenced in the text
• Citations formatted correctly
• Self-assessment form completed (form here)

Thesis Evaluation and Presentation

Upon completion:

• Complete a detailed self-assessment (evaluation form), clearly referencing sections of your thesis.
• Review the grade characterization PDF for guidance on what constitutes a high-quality thesis.
• Prepare your thesis presentation (live during a group meeting or as a recorded video; see examples here).

Feedback and Questions

I'm always eager for your feedback or questions. Reach out via:

• Email: firstname.lastname@aalto.fi
• LinkedIn: linkedin.com/in/aljung/
• YouTube: @alexjung111
• GitHub: Use issues or pull requests.