Step 1

The most important things come first. We discuss the main message, delivered by a scientific paper. We explore the first three elements of a scientific paper: the abstract, the highlights, and the keywords. The main message shall reveal itself through all elements of the paper. But we leave the rest of it for the next time. Namely, the title, introduction, problem statement, goal of the computational experiment, and conclusion are left behind. We select a part and exercise in the reconstruction of these three elements.

The seminar

1. The warm-up 3-minute test
2. Model, Algorithm, Method: Machine learning in a nut-shell
3. Step 0 homework results discussion
4. Step 1 homework, how to read (how to search is a separate topic)
5. Structure of the main message
6. Structure of the abstract
7. The second and the last slide of your talk
8. Extracting keywords
9. Highlights: compressing the paper
10. Instastructure for your homework
    • GitHub: organize the repository
    • LaTeX: compile your file and commit without temporary files
11. The papers to select from
12. Optional GPT-role discussion

Resources

Step 1 Youtube video (expected with online version)

Homework

1. Set up your GitHub repository using the template
2. Select a paper to read from the list below
3. Write your own
   1. Abstract
   2. Keywords
   3. Highlights
   4. Short motivation for why you selected this paper (no templates here it is an extra topic to discuss)
4. Compile and upload TEX and PDF to GitHub (no temporary files, please)
5. Linear models. Read pages XX from the book XX, for the next warm-up test

Note that we always respect your credit hours. So please keep track of it.

Papers to read

Can I select a paper of my own choice?

Yes. Here are some formal requirements.

1. A clear message in the area of Theoretical Foundations of Machine Learning.
2. No Kaggle-style papers with messages like "It works, but nobody knows how".
3. Top peer-reviewed journals, no ArXiv, better avoid conferences.
4. No papers from other fields: linguistics, medicine, finance, physics, etc. There must be only one primary subject: Machine Learning.
5. No overviews of paper collections, it is another genre.
6. No Predatory publishing houses

Recommended journals

1. Machine Learning
2. Expert Systems with Applications
3. Journal of Machine Learning Research
4. Artificial Intelligence
5. Neurocomputing
6. IEEE Transactions on Pattern Analysis and Machine Intelligence
7. Neural Networks
8. Pattern Recognition
9. Data Mining and Knowledge Discovery
10. Nature Machine Inlelligence, the problem is the first word here is Nature so it focuses on natural sciences

See also

1. Elsevier's open archive
2. Springer's open archive

Transcript of the video

Appears after the seminar.