Quick Reference: AI & ML Research Guide¶

⚡ Quick Reference

One-Page Overview of Research Essentials

Quick Overview

This is a condensed reference. See individual chapters for detailed information.

📚 Course Structure¶

Beginner Level (Chapters 1-5)¶

Introduction - What is research?
Reading Papers - Three-pass approach
Finding Resources - Databases and tools
Topic Selection - Choosing research topics
Basic Methodology - Research fundamentals

Intermediate Level (Chapters 6-10)¶

Literature Review - Comprehensive surveys
Research Design - Experimental setup
Data Management - Data handling
Reproducing Papers - Code reproduction
Research Tools - Essential tools

Advanced Level (Chapters 11-15)¶

Writing Papers - Paper writing
Publication - Submission process
Ethics - Research ethics
Advanced Topics - Cutting-edge areas
Career - Research careers

🔍 Paper Reading: Three-Pass Approach¶

First Pass (5-10 min)¶

Read title, abstract, introduction
Glance at figures and headings
Read conclusion
Goal: Determine relevance

Second Pass (30-60 min)¶

Read entire paper carefully
Take notes on key points
Understand main contributions
Goal: Understand methodology

Third Pass (1-2 hours)¶

Read every detail
Understand math and algorithms
Trace through experiments
Goal: Deep understanding

📖 Paper Structure¶

Title & Abstract
├── Introduction (motivation, contributions)
├── Related Work (existing methods)
├── Methodology (your approach)
├── Experiments (results, comparisons)
├── Results & Discussion (analysis)
└── Conclusion (summary, future work)

🎯 Research Process¶

graph LR
    A[Problem] --> B[Literature Review]
    B --> C[Hypothesis]
    C --> D[Design]
    D --> E[Implementation]
    E --> F[Experiments]
    F --> G[Analysis]
    G --> H[Writing]
    H --> I[Submission]

🔬 Experimental Design Checklist¶

Before Experiments¶

During Experiments¶

Random seeds set
Multiple runs (3-5)
Experiment tracking active
Results logged

After Experiments¶

Statistics computed (mean ± std)
Ablation studies done
Comparisons made
Results documented

📊 Essential Metrics¶

Task Type	Metrics
Classification	Accuracy, F1, AUC-ROC
Regression	MSE, MAE, R²
Ranking	NDCG, MAP, MRR
Generation	BLEU, ROUGE, FID

🛠️ Essential Tools¶

Experiment Tracking¶

Weights & Biases - Most popular
MLflow - Open source
TensorBoard - Visualization

Version Control¶

Git + GitHub - Code versioning
DVC - Data versioning

Writing¶

LaTeX + Overleaf - Paper writing
Zotero - Reference management

Compute¶

Google Colab - Free GPU
Kaggle - Free compute
AWS/GCP - Cloud compute

📝 Paper Writing Tips¶

Title¶

10-15 words
Include key method/contribution
Searchable keywords

Abstract (150-250 words)¶

Problem statement
Proposed method
Key results
Impact

Writing Style¶

✅ Clear and concise
✅ Active voice
✅ Simple words
❌ Verbose
❌ Vague
❌ Jargon

🎓 Top Conferences¶

ML Conferences¶

NeurIPS (Dec) - Neural Information Processing Systems
ICML (Jul) - International Conference on Machine Learning
ICLR (May) - International Conference on Learning Representations
AAAI (Feb) - Association for the Advancement of AI

CV Conferences¶

CVPR (Jun) - Computer Vision and Pattern Recognition
ICCV (Oct, biennial) - International Conference on Computer Vision
ECCV (Aug, biennial) - European Conference on Computer Vision

🔗 Essential Resources¶

Paper Discovery¶

arXiv - https://arxiv.org/list/cs.LG/recent
Papers With Code - https://paperswithcode.com/
Google Scholar - https://scholar.google.com/
Semantic Scholar - https://www.semanticscholar.org/
Connected Papers - https://www.connectedpapers.com/

Reference Management¶

Zotero - https://www.zotero.org/ (Free, recommended)
Mendeley - https://www.mendeley.com/

Writing Tools¶

Overleaf - https://www.overleaf.com/ (LaTeX editor)
Grammarly - https://www.grammarly.com/ (Writing assistant)

⚠️ Common Mistakes to Avoid¶

Experimental¶

❌ Data leakage (preprocessing before split)
❌ Tuning on test set
❌ Single runs (no statistics)
❌ Unfair comparisons
❌ Missing ablations

Writing¶

❌ Poor figures/tables
❌ Inadequate citations
❌ Missing limitations
❌ Unclear contributions
❌ Typos and errors

Research¶

❌ No reproducibility
❌ Cherry picking results
❌ Ignoring related work
❌ Poor organization
❌ No documentation

✅ Best Practices¶

Research¶

✅ Reproducible experiments
✅ Multiple runs with statistics
✅ Fair comparisons
✅ Ablation studies
✅ Honest reporting

Code¶

✅ Version control (Git)
✅ Clean, documented code
✅ Share code when possible
✅ Test thoroughly

Writing¶

✅ Clear and concise
✅ High-quality figures
✅ Proper citations
✅ Acknowledge limitations
✅ Proofread carefully

🎯 Research Topic Criteria¶

Good research topics are: - Novel: Addresses something new - Significant: Has potential impact - Feasible: Can be completed with resources - Interesting: Aligns with your interests - Clear: Well-defined and specific

📈 Research Career Paths¶

Academic¶

PhD Student → Postdoc → Assistant Prof → Associate/Full Prof

Industry¶

Research Scientist → Senior Scientist → Research Director

Timeline¶

PhD: 4-6 years
Postdoc: 2-4 years
Tenure: 6-7 years

🔑 Key Concepts¶

Concept	Definition
Baseline	Simple method for comparison
Ablation	Remove components to understand contribution
SOTA	State-of-the-art (best known method)
Reproducibility	Ability to reproduce results
Novelty	New contribution of work

💡 Quick Tips¶

Reading Papers¶

Start with abstract to decide relevance
Use three-pass approach
Take systematic notes
Build reading list

Experiments¶

Split data before preprocessing
Use validation set for tuning
Test set only for final evaluation
Multiple runs with statistics

Writing¶

Follow standard structure
Write clearly and concisely
Create high-quality figures
Cite properly
Proofread carefully

Career¶

Network actively
Build portfolio
Seek mentorship
Maintain balance

🚨 Red Flags in Papers¶

No baselines or weak baselines
Small datasets (may not generalize)
Missing implementation details
Overstated claims
Poor experimental design
No code available

📞 Quick Links¶

Course Overview: README
Chapter 1: Introduction
Chapter 2: Reading Papers
Chapter 11: Writing Papers
Chapter 12: Publication

Remember: Research is a journey. Start with fundamentals, build skills gradually, and stay persistent. Good luck with your research!

Last Updated: November 2024