Guide
Research guide
How the lab expects a study to be run, from choosing a problem to submitting the paper.
How to Select a Problem
Start by identifying your strengths and interests. Browse the Problem Board, filter by area and difficulty, and read the problem statements carefully. Choose a problem where you can contribute meaningfully while learning something new.
Literature Review
Before diving into solutions, read existing papers and approaches. Use Google Scholar, arXiv, and Semantic Scholar. Aim to read 15-20 relevant papers. Create a literature matrix comparing methods, datasets, and results.
Finding Research Gaps
Every problem in the Lab has a documented research gap. Your job is to validate and refine it. Look for assumptions in existing work that don't hold, datasets that are outdated, or evaluation metrics that miss important aspects.
Designing Experiments
Follow the scientific method. Formulate hypotheses, design controlled experiments, choose appropriate baselines, and define clear evaluation metrics. Document everything in a research log.
Writing a Paper
Structure: Abstract → Introduction → Related Work → Method → Experiments → Conclusion. Write the abstract last. Use clear, concise language. Every claim must be backed by evidence.
Git/GitHub Workflow
All research code lives on GitHub. Use feature branches, write descriptive commit messages, and open PRs for review. Tag releases with paper versions for reproducibility.
Research Ethics
Follow ethical guidelines for human subjects research, data privacy, and responsible disclosure. If your research involves datasets with personal information, ensure IRB compliance.
Publishing
Aim for workshops and conferences in your field. We support submission to arXiv for preprints. The Lab can provide feedback on manuscripts and help with submission logistics.