Hello everyone,
I’ve recently started reading ML research papers, and I’d love to hear your best practices, recommendations, tips on how to find a good research paper, and any personal favorites.
If anyone’s interested, here’s my current approach to reading ML research papers (open to improvements):
- Abstract – The abstract is the most important part, as it provides the relevance and main topic of the paper.
- Conclusion – This section helps clarify the paper’s direction, allowing me to approach the reading with a clear objective.
- Data Section – Understanding the data (X, Y) gives me an initial sense of how the model might work, even before delving into its structure.
- Results – This section reveals how the dataset performs and often includes performance comparisons.
- Dive In – Read the paper in detail, focusing on how the model is built:
- Model Architecture
- Inputs & Outputs
- New Techniques (usually highlighted in the title/abstract)
- Loss Calculation
- Model Training
- Note any unclear points
- Experiment – If possible, try running the code and dataset yourself.
This template helps me read scientific papers on ML, but it’s far from perfect since I’m adapting it as I go, based on limited experience.
Any recommendations would be VERY appreciated, as I’m still a beginner in the research field.
See you guys arround !!