Skill UI

This guide provides best practices for ensuring the scientific rigor and reproducibility of data mining and machine learning research published at top-tier conferences like KDD. It details a three-tier honesty model (Rerunnable, Rebuildable, Attested) and provides comprehensive checklists for disclosing pipeline stages, handling stochasticity, and documenting configurations to build reader trust.

A comprehensive guide for building robust causal identification strategies in empirical research, particularly for analyzing policy shocks. It provides best practices and checklist items for advanced econometrics methods like Difference-in-Differences (DID), Instrumental Variables (IV), Regression Discontinuity (RDD), and Double Machine Learning (DML), helping researchers avoid common methodological pitfalls in top-tier academic publications.

A comprehensive guide for authors submitting to machine learning conferences like NeurIPS. It details the rigorous standards for data splitting, hyperparameter reporting, random-seed protocols, and compute transparency required for a main-track paper. It also helps authors strategically decide between the main track and the MLRC/TMLR routes based on their core contribution (replication vs. novelty).

This guide outlines comprehensive best practices for ensuring the reproducibility of Machine Learning and Pattern Recognition research. It covers fixing random seeds, locking hardware/software dependencies (Python, CUDA, etc.), managing data splits, tracking pre-trained model versions, and structuring the required artifact package for external review. It is essential for building trust in published scientific results.

A comprehensive guide for researchers developing systems or algorithms, detailing the necessary level of evidence required for performance claims (e.g., bounds, real-world performance, learning guarantees). It emphasizes scientific rigor, statistical reporting (confidence intervals), assumption testing, and data provenance to ensure findings are robust and reproducible in academic settings.

This guide provides a comprehensive framework for designing rigorous and defensible evaluation sections for machine learning and information retrieval papers, particularly for top-tier conferences like WSDM. It covers the four critical evidence quadrants—Effectiveness, Validity, Efficiency, and Robustness—and details necessary protocols, such as temporal splits, proper baseline selection (including heuristics), ablation studies, and hardware-aware performance reporting. Use this when auditing or writing the experimental section of a research paper.

Expert data scientist for advanced analytics, machine learning, and statistical modeling. Guides the full workflow from exploratory data analysis to production deployment, covering experimentation, forecasting, NLP, and business intelligence for actionable insights.